American Journal of Botany 98(4): 704–730. 2011.

A NGIOSPERM PHYLOGENY: 17 GENES, 640 TAXA 1

Douglas E. Soltis2,19 , Stephen A. Smith3 , Nico Cellinese 4 , Kenneth J. Wurdack 5 , David C. Tank6 , Samuel F. Brockington 7 , Nancy F. Refulio-Rodriguez 8 , Jay B. Walker9 , Michael J. Moore 10 , Barbara S. Carlsward 11 , Charles D. Bell12 , Maribeth Latvis2,4 , Sunny Crawley13 , Chelsea Black 13 , Diaga Diouf13,14 , Zhenxiang Xi 15 , Catherine A. Rushworth15 , Matthew A. Gitzendanner2,4 , Kenneth J. Sytsma9 , Yin-Long Qiu16 , Khidir W. Hilu13 , Charles C. Davis15 , Michael J. Sanderson17 , Reed S. Beaman 4 , Richard G. Olmstead8 , Walter S. Judd2 , Michael J. Donoghue 18 , and Pamela S. Soltis4

2 Department of Biology, University of Florida, Gainesville, Florida 32611-8525 USA; 3 Department of Ecology and Evolutionary Biology, Brown University, Providence, Rhode Island 02912 USA; 4 Florida Museum of Natural History, University of Florida, Gainesville, Florida 32611-7800 USA; 5 Department of Botany, National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20013-7012 USA; 6 Department of Forest Ecology and Biogeosciences & Stillinger Herbarium, University of Idaho, P. O. Box 441133, Moscow, Idaho 83844-1133 USA; 7Department of Plant Sciences, University of Cambridge, Cambridge, CB2 3EA, UK; 8 Department of Biology, University of Washington, Seattle, Washington 98195-5325 USA; 9 Department of Botany, University of Wisconsin, Madison, Wisconsin 53706 USA; 10 Biology Department, Oberlin College, Oberlin, Ohio 44074-1097 USA; 11 Department of Biological Sciences, Eastern Illinois University, Charleston, Illinois 61920 USA; 12 Department of Biological Sciences, University of New Orleans, New Orleans, Louisiana 70148 USA; 13 Department of Biological Sciences, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061 USA; 14 D é partement de Biologie V é g é tale, Universit é Cheikh AntDeara Diop, Dakar – Fann, BP 5005, Republic of S é n é gal; 15 Department of Organismic and Evolutionary Biology, Harvard University Herbaria, 22 Divinity Avenue, Cambridge, Massachusetts 02138 USA; 16 Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, Michigan 48109-1048 USA; 17 Department of Ecology and Evolutionary Biology, University of Arizona, 1041 East Lowell, Tucson, Arizona 85721-0088 USA; and 18 Department of Ecology and Evolutionary Biology, Yale University, P. O. Box 208106, New Haven, Connecticut 06520-8106 USA

• Premise of the study: Recent analyses employing up to fi ve genes have provided numerous insights into angiosperm phylogeny, but many relationships have remained unresolved or poorly supported. In the hope of improving our understanding of angio- sperm phylogeny, we expanded sampling of taxa and genes beyond previous analyses. • Methods: We conducted two primary analyses based on 640 species representing 330 families. The fi rst included 25 260 aligned base pairs (bp) from 17 genes (representing all three plant genomes, i.e., nucleus, plastid, and mitochondrion). The second included 19 846 aligned bp from 13 genes (representing only the nucleus and plastid). • Key results : Many important questions of deep-level relationships in the nonmonocot angiosperms have now been resolved with strong support. Amborellaceae, Nymphaeales, and Austrobaileyales are successive sisters to the remaining angiosperms (Mesangiospermae ), which are resolved into Chloranthales + Magnoliidae as sister to Monocotyledoneae + [Ceratophyllaceae + Eudicotyledoneae ]. Eudicotyledoneae contains a basal grade subtending Gunneridae . Within Gunneridae, Gunnerales are sis- ter to the remainder (Pentapetalae ), which comprises (1) Superrosidae , consisting of Rosidae (including Vitaceae) and Saxi- fragales; and (2) Superasteridae , comprising Berberidopsidales, Santalales, Caryophyllales , Asteridae , and, based on this study, Dilleniaceae (although other recent analyses disagree with this placement). Within the major subclades of Pentapetalae , most deep-level relationships are resolved with strong support. • Conclusions: Our analyses confi rm that with large amounts of sequence data, most deep-level relationships within the angio- sperms can be resolved. We anticipate that this well-resolved angiosperm tree will be of broad utility for many areas of biology, including physiology, ecology, paleobiology, and genomics.

Key words: angiosperms; bioinformatics; large data sets; molecular systematics; RAxML; Superasteridae ; supermatrix; Superrosidae .

1 Manuscript received 11 October 2010; revision accepted 9 February 2011. C.C.D., DEB-9634231; EF-0431057 to K.H., EF-0431233 to K.J.S., The authors thank the University of Florida High-Performance DEB-0431239 to Y.L.Q.) with additional support from NSF grant DEB- Computing Center, the UF Genetics Institute, and facilities supported in 0544039 to C.C.D. The authors thank the Royal Botanic Gardens, Kew, part by NSF Grant CNS-0821622 for providing computational resources and the Missouri Botanical Garden for providing samples used in this that facilitated the completion of this research. The authors thank E. study. Mavrodiev for help in preparing some of the fi gures. This study was 19 Author for correspondence (e-mail: [email protected]fl .edu) carried out as part of the Angiosperm Tree of Life Project (NSF EF- 0431266 to D.E.S., P.S.S., W.S.J., and S. R. Manchester, EF-0431242 to doi:10.3732/ajb.1000404

American Journal of Botany 98(4): 704–730, 2011; http://www.amjbot.org/ © 2011 Botanical Society of America 704 April 2011] Soltis et al. — Angiosperm phylogeny 705

It has been a decade since the 567-taxon, three-gene (rbcL, MATERIALS AND METHODS atpB, and 18S rDNA) parsimony-based phylogenetic analysis of angiosperms was published (Soltis et al., 1999, 2000 ). Since Data sets— DNA samples for most of the species used here were extracted that time, other studies have expanded sampling of additional from either fresh or silica-dried material following the general method of Doyle genes or used alternative methods to evaluate the results of that and Doyle (1987) or modifi cations thereof that employ liquid nitrogen and higher CTAB concentrations (e.g., Soltis et al., 1991; Sytsma, 1994). We at- three-gene study. Hilu et al. (2003) conducted a broad analysis tempted to use the same species and DNA samples across all of the genes ana- of angiosperms based on matK sequences, with results that lyzed here, although multiple species were sometimes used as necessary agreed closely with the three-gene topology. Additional prog- placeholders to reduce missing data. Many of these DNA samples have been ress was achieved by Davies et al. (2004) by constructing a used in earlier analyses (e.g., Chase et al., 1993 ; Soltis et al., 2000 ). supertree for angiosperms. Soltis et al. (2007) undertook a We constructed a 17-gene data set for 640 species (for a complete list of taxa, Bayesian analysis of the 567-taxon, three-gene data set and ob- voucher information, and GenBank numbers see Appendix S1, see Supple- mental Data online at http://www.amjbot.org/cgi/content/full/ajb.1000404/DC1) tained a topology nearly identical to that obtained with parsi- using genes from the nuclear, plastid, and mitochondrial genomes. Given mony. More recently, Bell et al. (2010) analyzed this same the potential problems inherent in phylogeny reconstruction using mtDNA se- three-gene data set using a Bayesian relaxed clock model to si- quences (e.g., RNA editing, Bowe and dePamphilis, 1996 ; horizontal gene multaneously infer topology and divergence times within an- transfer or HGT, Bergthorsson et al., 2003 ), we also constructed a data set giosperms and found results similar to Soltis et al. (2007). In without the mtDNA data, resulting in a matrix of 13 genes. For both data sets, addition, Burleigh et al. (2009) inferred angiosperm phylogeny we used the following representatives of Acrogymnospermae (extant gymno- sperms, sensu Cantino et al., 2007 ) as the outgroup: Cycas , Ginkgo , Gnetum , using fi ve genes for the same 567 taxa analyzed in Soltis et al. Metasequoia , Pinus , Podocarpus , Welwitschia, and Zamia. Taxon sampling (1999, 2000 ). Although the fi ve-gene matrix had signifi cantly across major clades of angiosperms was not uniform, with poorly resolved more missing data (27.5%) than the three-gene matrix (2.9%), clades (e.g., Malpighiales, Saxifragales) targeted for denser taxon sampling to the fi ve-gene analysis resulted in higher levels of bootstrap sup- seek improved resolution. We also largely avoided parasitic clades (except port across the tree. Orobanchaceae, Santalales, and Cuscuta), which can create analytical prob- lems due to gene loss, accelerated molecular evolution, and horizontal gene Recent phylogenomic analyses have shown the value of con- transfer (see Davis and Wurdack, 2004 ; Nickrent et al., 2004 ; Barkman et al., structing data sets of many genes to infer deep-level angiosperm 2007 ). phylogeny. Some of these analyses have employed nearly com- The following 17 genes were sequenced: 18S and 26S rDNA from the plete plastid genome sequence data (e.g., Leebens-Mack et al., nuclear genome; atpB , matK, ndhF, psbBTNH (four contiguous genes here treated 2005; Jansen et al., 2007; Moore et al., 2007, 2010 ), but all as one region), rbcL , rpoC2 , rps16 , and rps4 from the plastid genome; and atp1 , have been limited in sampling to fewer than 100 taxa. These matR , nad5 , and rps3 from the mitochondrial genome. The total length of the aligned 17-gene matrix was 25 260 bp and of the 13-gene matrix was 19 846 bp. and other studies based on many genes but focused only on The percentage of missing data for the full data set was 41% and for the data set major angiosperm clades (e.g., Sch ö nenberger et al., 2005; Jian without mtDNA data was 42%. et al., 2008 ; H. Wang et al., 2009 ; Wurdack and Davis, 2009 ; Brockington et al., 2010; Tank and Donoghue, 2010) showed Alignment and phylogenetic analyses —All sequence data were stored and that with very large amounts of data (i.e., 13 to 83 genes), many, managed in TOLKIN ( Beaman and Cellinese, 2010 ). TOLKIN is a web appli- if not most, deep-level questions of angiosperm phylogeny can cation, developed for distance collaboration as part of the Angiosperm Tree of be resolved. Life project, that allows users to access and share data in real time, as well as While the three- and fi ve-gene analyses of 567 taxa have automatically generate FASTA fi les and link to other relevant information (e.g., taxonomy and vouchers) and resources (e.g., GenBank, TreeBASE, and uBIO). broad taxonomic coverage, support for many portions of the The sequences generated here were supplemented with those already available framework of angiosperm phylogeny is low in these studies. in GenBank to obtain a more complete data set. GenBank sequences were re- For example, relationships among members of Mesangiospermae trieved using the PHLAWD package ( Smith et al., 2009 ; http://code.google. and of Pentapetalae remain unclear. Conversely, studies em- com/p/phlawd), and alignments of combined sequences were generated with ploying complete plastid genome sequences have deep gene the program MAFFT (vers. 6.71; Katoh and Toh, 2008 ) at the DNA level using coverage and strong internal support, but taxonomic coverage the l-ins-i algorithm and default alignment parameters. MAFFT was chosen because of its strong performance over a range of alignment scenarios ( Golubchik is often sparse. Hence, it is important to assemble a data set et al., 2007 ). Subsequent adjustments were made by eye when there were obvi- having both broad taxonomic coverage as well as numerous ous alignment errors due to particularly divergent or “ gappy ” sequences. The genes. In the hope of further improving our understanding of individual gene regions varied in the amount of missing data per site: 18S angiosperm phylogeny, we have sequenced deeply over a broad rDNA (6%), 26S rDNA (15%), atpB (5%), atp1 (1%), matK (13%), matR (3%), representation of angiosperms. We constructed a 17-gene data nad5 (4%), ndhF (20%), psbBTNH (19%), rbcL (4%), rpoC2 (21%), rps16 (26%), rps3 (10%), and rps4 (50%). Individual gene regions also varied in the set for 640 species representing 640 genera, 330 families, and number of taxa with data in the combined analyses: 18S rDNA (78%), 26S 58 of 59 orders (sensu APG III, 2009) using genes that repre- rDNA (57%), atpB (88%), atp1 (59%), matK (92%), matR (76%), nad5 (59%), sent all three plant genomic compartments (nucleus, plastid, ndhF (80%), psbBTNH (54%), rbcL (98%), rpoC2 (63%), rps16 (35%), rps3 and mitochondrion). (62%), and rps4 (58%). Sites in the alignment with more than 50% missing One goal of the angiosperm Assembling the Tree of Life data were removed with the program Phyutility (Smith and Dunn, 2008; (AToL) project was to assess morphological synapomorphies see discussion in Castresana, 2000 ) to avoid regions of potentially problem- atic ambiguous alignment caused by such broad sampling ( Talavera and for each of the major clades of angiosperms as well as across Castresana, 2007 ). angiosperms as a whole. We here also provide putative mor- Phylogenetic analyses using maximum likelihood ( Felsenstein, 1973 ) were phological synapomorphies of Saxifragales, although a more conducted in the program RAxML (vers. 7.1; Stamatakis, 2006 ). For each data detailed morphological exploration of this clade is presented set, we searched for the optimal tree, running at least 10 independent maximum elsewhere ( Carlsward et al., unpublished manuscript ). This likelihood analyses; full analyses also consisted of at least 100 and up to 300 study of Saxifragales represents the fi rst treatment in a series bootstrap replicates (Stamatakis et al., 2008). We conducted analyses on all individual genes, the concatenated 17-gene data set, the concatenated 13-gene of analyses for large clades of angiosperms that will address data set (no mtDNA data), and genomic compartments (nucleus, plastid, and morphological synapomorphies and the clades they support mitochondrion). The GTRGAMMA substitution model was applied to (W. S. Judd et al., unpublished data). each gene independently. For analyses of all concatenated data sets, all genes 706 American Journal of Botany [Vol. 98 were partitioned, and unlinked substitution models were applied to each gene. The 17-gene tree (with mtDNA for all taxa, including Poly- Bootstrapped (BS; Felsenstein, 1985 ) trees were summarized as majority-rule osma) and the 13-gene tree (without mtDNA) are presented as consensus trees with Phyutility ( Smith and Dunn, 2008 ). online supplementary fi gures (Appendix S2 and Appendix S3, A series of maximum likelihood (ML) analyses was conducted on the com- bined data sets, ultimately requiring approximately over 1 year of analysis time respectively). Additional online supplementary fi gures are pro- and approximately 9 years of actual CPU time. Unexpected taxon placements vided for the trees resulting from the analysis of each genomic required close examination of the alignment and subsequent reanalysis when partition: nuclear rDNA tree (Appendix S4), plastid tree (Ap- problems were detected. For example, initial analyses indicated that the place- pendix S5), and mtDNA tree (Appendix S6), as well as the ment of Polyosma (Escalloniaceae) differed dramatically when mtDNA data single-gene trees (Appendices S7– S20; note that psbBNTH , were included vs. when they were omitted for this taxon (see Results). Parti- representing four genes, are combined in one tree). We also tioned analyses suggested that the mtDNA signal for Polyosma was providing a spurious placement. In our total evidence analysis (Kluge, 1989; Fig. 1, sum- provide a tree resulting from analysis of a reduced, 17-gene, mary tree, and Fig. 2 , complete cladogram; shown as a phylogram in Appendix 86-taxon data set (Appendix S21) for comparison with a recent S1), we therefore omitted the mtDNA data for Polyosma . We subsequently se- 86-taxon tree based on complete plastid genome sequences quenced the mitochondrial genes surveyed here for an additional sample of ( Moore et al., 2010 ). Polyosma as well as for a second accession of Quintinia , the sister of Polyosma Finally, we also provide an MP total evidence topology, with in the trees that included mtDNA sequence data. mtDNA for all taxa, except Polyosma (Appendix S22). As re- Maximum parsimony (MP) has not performed well on some recent large angiosperm data sets; long-branch attraction has clearly played a role in previ- viewed below, the MP topology (Appendix S22) is very similar ous analyses of deep-level relationships in angiosperms, especially in analyses to the ML tree ( Figs. 1, 2 ). with limited taxon sampling (reviewed in Leebens-Mack et al., 2005 ; Soltis Taxon sampling is so dense, it is decisive for all possible to- et al., 2005, 2007 ; Burleigh et al., 2009). Given the current availability of ML pologies in the sense of Steel and Sanderson (2010 ; see also programs (e.g., RAxML) that can readily handle large data sets of the size em- Sanderson et al., 2010). This means that there is suffi cient taxon ployed here, we focused much of our data analyses on this approach. Nonethe- coverage to avoid problems of lack of resolution due solely to less, we also conducted MP searches, using the parsimony ratchet (Nixon, 1999) approach to thorough and rapid tree searching. These analyses were run missing data (though lack of resolution stemming from insuffi - in the program PAUP* 4.0b10 (Swofford, 2002). Ratchet fi les were generated cient sequence information may still be an issue). with PAUPRat ( Sikes and Lewis, 2001 ) with 50 independent replicates of 500 iterations each. A majority-rule consensus of the best trees from each replicate was generated. Bootstrapping was conducted by generating 500 bootstrap data DISCUSSION sets with the SeqBoot module of PHYLIP ( Felsenstein, 2005 ) and running each of these with a PAUPRat-generated ratchet fi le for a single 500-iteration search. Overview— Broad phylogenetic analyses involving three to The parsimony searches took just over three CPU years of analysis time. It is fi ve genes have provided important insights into angiosperm noteworthy therefore that, for this data set, a thorough MP analysis took longer than a comparable ML analysis (each total evidence ML analysis required one phylogeny, but crucial portions of the backbone of the tree were CPU year of analysis time). either not resolved or not well supported (e.g., relationships For higher clades, we consistently use PhyloCode names (see Cantino et al., among major lineages of Mesangiospermae and of Pentapeta- 2007) whenever these are available; these names are always in italics (e.g., lae). Use of large numbers of genes from nearly compete plas- Pentapetalae , Mesangiopsermae , Rosidae , Fabidae , Malvidae ). Note that Rosi- tid genome sequences established most deep branches of extant dae (sensu Cantino et al., 2007) does include Vitaceae. Our use of family and angiosperm diversity with strong support, but taxon density ordinal names follows APG III (2009) as a formal point of reference; for Caryo- phyllales, we follow Cantino et al. (2007; hence, the use of italics), which was low ( Jansen et al., 2007 ; Moore et al., 2007 , 2010 ). Here, matches the APG III circumscription. For additional recent discussion on fami- we have employed numerous genes as well as broad taxonomic lies and their status, see the Angiosperm Phylogeny Website ( Stevens, 2001 coverage. The general topology provided here is very similar to onward). We recognize that some broader family circumscriptions favored in recently published trees that include broad taxonomic coverage APG III are controversial and can obscure underlying diversity (e.g., Passifl o- of the angiosperms, albeit with fewer genes (e.g., Soltis et al., raceae s.l.), which would be evident with narrower circumscriptions. All align- 2000, 2007 ; Burleigh et al., 2009) or far fewer taxa and nearly ments and trees have been deposited in TreeBASE (no. 11267; see http://www. complete plastid genome sequences (e.g., Jansen et al., 2007; treebase.org/). After all analyses were complete, the possibility was raised that the atp1 Moore et al., 2007, 2010 ). However, the 17-gene tree is a sig- sequence of Cardiopteris was a contaminant. This sequence did not impact the nifi cant improvement in that it provides much higher levels of fi nal placement of the genus, but as a result the questionable sequence was not support for deep-level relationships than obtained with either submitted to GenBank. three or fi ve genes. Our analyses further support those topolo- gies recovered using complete plastid genome analyses (but based on fewer than 100 exemplars). RESULTS Within Acrogymnospermae , the two Gnetales included here ( Gnetum and Welwitschia ) are embedded within the Coniferae Each ML analysis of the 17- and 13-gene data sets took 20– (conifers) and are sister to Pinaceae (BS = 81%) in any poten- 32 h on a 32-core (2.93 gHz Xeon × 7350) machine with 128 gb tial rooting other than one in which Gnetales are sister to all RAM, and analyses of individual genes took 1 – 11 h. other seed plants. The best RAxML trees from analyses of the 17-gene and 13- Within the angiosperms, Amborellaceae, Nymphaeales, and gene matrices are very similar, but with a few noteworthy dif- Austrobaileyales are subsequent sisters to all other extant fl ow- ferences. The 17-gene tree has one major deviation in placement ering plants. These placements all receive BS support (BS) from expected relationships (Polyosma ; see above). However, > 80%; Fig. 1) and are a result obtained across 17-gene and 13- other differences are relatively minor, and in most cases the 17- gene data sets. It is noteworthy, however, of the genome partitions, gene tree gives higher BS support than the 13-gene tree. Hence, the mtDNA data have Amborella + Nymphaeales (BS = 88%), the only the 17-gene tree (total evidence, but no mtDNA data for cpDNA partition has Amborella + Nymphaeales (BS = 78%), and Polyosma; e.g., Lecointre and Deleporte, 2005) is discussed be- the 18S/26S rDNA partition has Nymphaeales followed by low. This tree (Fig. 1, summary tree) has been divided into sep- Amborella as sister to other angiosperms (BS < 50%; Appendices arate, interconnected subtrees ( Fig. 2a-l ). S4 – S6). Our interpretation of these results is that the support April 2011] Soltis et al. — Angiosperm phylogeny 707

Fig. 1. Summary tree of the maximum likelihood majority-rule consensus from the 17-gene analysis with mtDNA data removed for Polyosma . Names of the orders and families follow APG III (2009) ; other names follow Cantino et al. (2007) . Numbers above branches are bootstrap percentages. The number and letter following clade names (i.e., 2a to 2l) refer to clade designations that are used to depict the separate portions of the complete tree in Fig. 2 . 708 American Journal of Botany [Vol. 98 from this particular data set for the placement of Amborella is Austrobaileyales, Austrobaileyaceae are sister to Trimeniaceae not strong. Nonetheless, the placement of Amborella as sister to + Schisandraceae (including Illiciacae; see APG III, 2009 ). all other extant angiosperms is well supported using complete plastid genome sequences (e.g., Leebens-Mack et al., 2005; Magnoliidae + Chloranthales— Chloranthaceae (the only Jansen et al., 2007 ; Moore et al., 2007 ), as well as in recent member of Chloranthales, APG III, 2009) are well supported analyses of numerous nuclear gene trees ( Jiao et al., in press ). here as sister to Magnoliidae (BS = 85%). This same sister- Following Amborellaceae, Nymphaeales, and Austrobailey- group relationship also emerged (with BS = 72%) from com- ales, the monophyly of the remaining angiosperms (Mesan- plete plastid genome sequencing ( Moore et al., 2010 ), but has giospermae) received maximum support. Chloranthales are not been apparent in previous analyses involving three or fi ve sister to the magnoliid clade ( Magnoliidae) with BS = 85%. genes (Soltis et al., 2000; Burleigh et al., 2009). Within Chlo- This clade of Magnoliidae + Chloranthales is in turn sister to ranthaceae, the relationships among the four genera [Hedyos- the remaining angiosperms. Monocotyledoneae (monocots), mum , (Ascarina (Sarcandra + Chloranthus )] agree with the Ceratophyllaceae, and Eudicotyledoneae (eudicots) form a results of focused analyses of the family (Qiu et al., 1999; Doyle well-supported clade (BS = 86%) with Monocotyledoneae sis- et al., 2003 ; Zhang and Renner, 2003 ; Eklund et al., 2004 ). ter to a weakly supported clade (BS = 68%) of Ceratophyllaceae + In agreement with other recent analyses (e.g., Zanis et al., Eudicotyledoneae. These deep-level relationships agree with 2002 ; Hilu et al., 2003 ; Qiu et al., 1999 , 2005 ; Soltis et al., those obtained using complete plastid genome sequences (Moore 2007 ; Moore et al., 2010 ), Magnoliidae comprises two well- et al., 2007 ). supported clades, each with BS = 100%: Magnoliales + Lau- Within Eudicotyledoneae , there is a grade of basal taxa; in rales and Piperales + Canellales. Relationships within these contrast to many previous studies, relationships among the four clades also agree, for the most part, with previous analyses members of this basal grade are well supported. Successively and are summarized below. more distant sister groups from Gunneridae (= core eudicots) Piperales consist of two well-supported clades: Piperaceae + are Buxaceae, Trochodendraceae, Proteales plus Sabiaceae, Saururaceae (BS = 100%) as sister to Aristolochiaceae + Lacto- and Ranunculales. ridaceae (BS = 99%). The parasitic Hydnoraceae, also of Piper- Within Gunneridae, Gunnerales are sister to the remainder of ales ( APG III, 2009 ), were not included here. Canellales this clade (BS = 99%), which constitute Pentapetalae. Penta- comprise Canellaceae + Winteraceae. petalae comprises (1) a well-supported (BS = 100%) “ super- In Magnoliales, Magnoliaceae are sister to the remaining rosid ” clade, here named Superrosidae (see below), consisting Magnoliales: Degeneriaceae + Myristicaceae are sister to Him- of Rosidae (including Vitaceae) and Saxifragales; and (2) a antandraceae + [Eupomatiaceae + Annonaceae], albeit without well-supported (BS = 87%) “ super-asterid ” clade, here named BS > 50%. These results differ from the three-gene ( Soltis et al., Superasteridae (see below), consisting of Berberidopsidales, 2000 , 2007 ) and fi ve-gene (Burleigh et al., 2009) analyses, as Santalales, Caryophyllales , Asteridae, and Dilleniaceae, whose well as focused studies on the clade (Sauquet et al., 2003), all of position here is at odds with other recent studies (see below). which have placed Myristicaceae as sister to all other Magno- The MP topology (Appendix S22) is similar to the ML tree liales. The problem with the placement of Myristicaceae here (Figs. 1, 2), but there are several noteworthy differences that appears to be the result of the rDNA data. Both plastid and have also been reported in previous analyses involving MP and mtDNA place Myristicaceae as sister to all remaining Magno- ML on smaller data sets (see Soltis et al., 2005 , 2007 ). For ex- liales, following earlier analyses. However, the combined 18S + ample, in the MP tree, Ceratophyllaceae are sister to the Mono- 26S rDNA tree places a strongly supported clade (BS = 94%) cotyledoneae rather than Eudicotyledoneae , as found with ML. of Myristicaceae plus Degeneriaceae (all with very long MP does recover the Superasteridae and Superrosidae clades, branches) as embedded well within Magnoliales, sister to An- but there are differences in relationships among members of nonaceae (Appendix S4). This result was not apparent in earlier Superasteridae with MP and ML. For example, with MP, Dil- studies that included 18S rDNA (Soltis et al., 2000). However, leniaceae are sister to Caryophyllales (rather than sister to all in this study, 18S rDNA placed Myristicaceae with Chloran- other Superasteridae , as found with ML). We will not discuss thaceae (albeit without BS support > 50%), and the 26S rDNA the MP topology further. data placed the family as indicated for the combined data set with strong BS support. Basal angiosperms— Our 17-gene analysis (Figs. 1, 2) places Within Laurales, Calycanthaceae are well supported as sister Amborellaceae, followed by Nymphaeales, and then Austrobai- to the remainder of the clade, as in previous analyses (e.g., Qiu leyales as well-supported sisters to all other extant angiosperms, et al., 1999; Soltis et al., 1999, 2000 ). The current analyses also in agreement with a series of phylogenetic analyses that have provide more resolution and support of relationships within the been based on an ever-increasing number of gene sequences remainder of Laurales than the three- and fi ve-gene studies (reviewed in Leebens-Mack et al., 2005 ; Soltis et al., 2005 ). ( Soltis et al., 2000 ; Burleigh et al., 2009 ), although reasonable Our results also agree with other analyses in providing strong resolution and support were obtained with six genes and mor- support (BS = 100%) for the placement of Hydatellaceae in phology (e.g., Renner, 1999). Our trees then divide the re- Nymphaeales as sister to other members of the clade ( Saarela maining Laurales into two subclades: (1) Lauraceae sister to et al., 2007). Trithuria has ascidiate carpel development, consis- (Hernandiaceae + Monimiaceae), and (2) Siparunaceae sister tent with placement in Nymphaeales, but as reviewed else- to (Atherospermataceae + Gomortegaceae). These same two where, this placement is important in that it greatly expands the clades were recovered by Renner (1999) , although the relation- morphological diversity encompassed by Nymphaeales (Rudall ships found here within the two clades are not identical to those et al., 2007 ). recovered in that study. Following Amborella and Nymphaeales, an Austrobaileyales clade (BS = 100%) of Schisandraceae, Austrobaileyaceae, and Monocotyledoneae— The monocots, or Monocotyledoneae , Trimeniaceae is sister to all remaining angiosperms. Within are not extensively sampled here, in deference to the ongoing April 2011] Soltis et al. — Angiosperm phylogeny 709

Fig. 2. The maximum likelihood majority-rule consensus from the 17-gene analysis shown as a cladogram with mtDNA data removed for Polyosma . This single large tree has been divided into a series of interconnected trees in which subclades are labeled 2a through 2l; these designations match those given in Fig. 1. Names of the orders and families follow APG III (2009); other names follow Cantino et al. (2007). Numbers above branches are bootstrap percentages. 710 American Journal of Botany [Vol. 98

Fig. 2. Continued. work of the Monocot AToL research group (see Givnish et al., lowed by Alismatales, each well supported (BS = 100%) as 2010) . Nonetheless, the relationships depicted here within the subsequent sisters to all other monocots (Petrosaviales were not Monocotyledoneae mirror those produced in multigene phylo- sampled). A clade of Dioscoreales + Pandanales (BS = 95%) is genetic analyses focused on the clade (e.g., Chase et al., 2006; sister to the remaining Monocotyledoneae . Within this remain- Graham et al., 2006 ). On a broad scale, Monocotyledoneae is der, Liliales followed by Asparagales are sisters to Commelinidae well supported (BS = 100%), with Acoraceae (Acorales), fol- (i.e., commelinid clade; BS = 100%; Arecales, Zingiberales, April 2011] Soltis et al. — Angiosperm phylogeny 711

Fig. 2. Continued.

Arecales, Poales, and Dasypogonaceae, a family not included sister to the rest of the clade. The remaining Ranunculales form here). Relationships within Commelinidae also agree with focused two clades: (1) Circaeasteraceae + Lardizabalaceae (BS = 78%), analyses of Monocotyledoneae , with Arecales sister to Com- and (2) Menispermaceae sister to (Berberidaceae + Ranuncu- melinales + Zingiberales; this clade is in turn sister to Poales. laceae) (BS = 100%). These results are comparable to most re- Relationships obtained here within the larger Monocotyledon- cent analyses (e.g., Kim et al., 2004 ; W. Wang et al., 2009 ), eae clades also largely agree with more focused analyses that although there has been disagreement in terms of the sister have included more taxa (see Chase et al., 2006; Graham et al., group to all other Ranunculales. In some analyses, Papaver- 2006 ; Givnish et al., 2010 ). aceae have appeared in this position (Soltis et al., 2000), whereas in other studies it has been Eupteleaceae (Kim et al., 2004; Eudicotyledoneae— Basal eudicots— Basal eudicot relation- W. Wang et al., 2009 ). ships parallel those in other broad analyses (e.g., Soltis et al., Within Proteales, Nelumbonaceae are sister to Platanaceae + 1999, 2000 ; Burleigh et al., 2009), albeit with higher BS sup- Proteaceae (with BS = 100%) as in other molecular analyses; port here. Ranunculales (BS = 100%) are sister to remaining recent investigations of fl oral morphology also indicate simi- Eudicotyledoneae, which form a clade with strong support (BS = larities between Proteaceae and Platanaceae ( von Balthazar and 100%); Ranunculales are then followed by a clade of Sabi- Sch ö nenberger, 2009 ). aceae + Proteales (BS = 59% in total evidence tree; 65% with- out mtDNA data); a similar sister-group relationship was Gunneridae— Gunnerales are sister to all other core eudicots recovered with strong support (BS = 80%) with complete plas- (i.e., Pentapetalae) with BS = 100%. Pentapetalae, in turn, tid genome data ( Moore et al., 2010 ). Following Proteales, both comprises: (1) the Superrosidae , consisting of Rosidae (in- analyses provided strong support for the placements of Trocho- cluding Vitaceae) and Saxifragales (BS = 100%); and (2) the dendraceae (BS = 100%) followed by Buxaceae (BS = 98%), as Superasteridae, consisting of Berberidopsidales, Santalales, subsequent sisters to the core eudicots (Gunneridae ). The same Caryophyllales , Asteridae, and possibly Dilleniaceae (BS = relationships were recovered by Moore et al. (2010) . 97%). Relationships within the basal eudicot clades are generally well resolved and supported. Within Ranunculales, Eupte- Superrosidae— Superrosidae consists of Saxifragales as leaceae are sister to the remaining taxa (BS = 76%), followed sister to a well-supported (BS = 85%) Rosidae ; Rosidae by Papaveraceae, which are well supported (BS = 100%) as comprises Vitaceae as sister to Malvidae + Fabidae (see 712 American Journal of Botany [Vol. 98

Fig. 2. Continued. April 2011] Soltis et al. — Angiosperm phylogeny 713

Fig. 2. Continued. 714 American Journal of Botany [Vol. 98

Fig. 2. Continued. April 2011] Soltis et al. — Angiosperm phylogeny 715

Fig. 2. Continued. 716 American Journal of Botany [Vol. 98

Fig. 2. Continued.

Cantino et al., 2007). Complete plastid genome data (Moore and form two major clades: (1) a weakly supported clade et al., 2010) as well as inverted repeat (IR) sequence data (Moore (BS = 59%) of Paeoniaceae + the “ woody clade ” (BS = 100%), et al., in press) also recover Superrosidae with strong support, comprising Cercidiphyllaceae, Daphniphyllaceae, Altingiaceae, but relationships differ among the three constituent clades (Vi- and Hamamelidaceae, and (2) core Saxifragales (BS = 100%), taceae, all remaining Rosidae , Saxifragales). Complete plastid comprising two clades, each with BS = 100%: Crassulaceae + genome data suggest that Rosidae (excluding Vitaceae) are sis- Haloragaceae s.l. (i.e., Aphanopetalum, Tetracarpaeaceae, ter to Saxifragales + Vitaceae, whereas IR sequence data re- Penthoraceae, and Haloragaceae) and the Saxifragaceae alli- cover the topology found here (Vitaceae sister to all remaining ance. Within the latter, Saxifragaceae + Grossulariaceae are Rosidae + Saxifragales). Focused studies of Superrosidae sug- sister to Iteaceae + Pterostemonaceae. These results agree with gest Saxifragales are sister to Vitaceae + all remaining Rosidae , the focused analyses of Jian et al. (2008) . in agreement with the results here (see H. Wang et al., 2009 ). Morphological synapomorphies for Saxifragales include The divergent results produced by analyses of complete plastid basifi xed anthers, usually with the fi lament attached at a basal genomes may be due to more limited taxon sampling in that pit (but reversals to the dorsifi xed condition occur in Ribes , analysis compared to the IR analysis (244 taxa; Moore et al., in Pterostemon, and Iteaceae), and latrorse dehiscence. All but Pe- press ) and the present study. ridiscaceae also may be united by follicle fruits, a homoplasious We provide a formal defi nition of Superrosidae in Appendix 1. character that is correlated with the ovaries being at least dis- tally distinct. Additionally, the presence of violoid to theoid Saxifragales— In the 17-gene tree, Peridiscaceae are sister to teeth is possibly synapomorphic, but salicoid teeth occur in all remaining Saxifragales, which are well supported (BS = 98%) Aphanopetalum , and nonglandular teeth occur in Hamamelis , April 2011] Soltis et al. — Angiosperm phylogeny 717

Fig. 2. Continued.

Itea , and Haloragis , while the condition cannot be assessed in largest poorly resolved major clade; deep relationships within taxa with entire leaves. Rosidae have been particularly problematic. Our analyses agree with recent analyses of this clade based on parsimony and ML Rosidae— Despite extensive progress in elucidating relation- analyses of 12-gene (> 20 000 bp) and IR (> 24 000 bp) data sets ships within the angiosperms, Rosidae has long stood out as the for over 100 rosid species ( H. Wang et al., 2009 ). Rosidae forms 718 American Journal of Botany [Vol. 98

Fig. 2. Continued. a well-supported clade (BS = 85%), in which Vitaceae are sister topologies resulting from analyses focused on Rosidae (H. Wang to the remainder (BS = 100%), which in turn consists of two et al., 2009). large clades: (1) Fabidae , which includes the nitrogen-fi xing Over half of the families of Brassicales (Akaniaceae, Em- clade, Celastrales, Huaceae, Zygophyllales, Malpighiales, and blingiaceae, Gyrostemonaceae, Koeberliniaceae, Moringaceae, Oxalidales; and (2) Malvidae , which includes Tapisciaceae, Pentadiplandraceae, Salvadoraceae, Setchellanthaceae, and To- Brassicales, Malvales, Sapindales, Geraniales, Myrtales, Cros- variaceae) were not included in our study. Of those Brassicales sosomatales, and Picramniaceae. sampled, relationships agree closely with other analyses (see Rodman et al., 1996, 1998 ; Hall et al., 2004). Tropaeolaceae Malvidae —Within a well-supported Malvidae (BS = 97%), and Caricaceae are sisters to a well-supported Limnanthaceae a clade (BS = 79%) of Myrtales (BS = 100%) + Geraniales followed by Bataceae, as subsequent sisters to Resedaceae + (BS = 68%) is sister to the remaining Malvidae (BS = 99%). (Capparidaceae + Brassicaceae). Malvales are also poorly sam- Within the latter, a well-supported (BS = 100%) Crossoso- pled, with only 5 of 10 families included, but the topology matales are sister to the rest (BS = 100%), followed by Picram- matches other studies (e.g., Bayer et al., 1999; Alverson et al., niaceae as sister to a clade (BS = 100%) of Sapindales 1999 ; H. Wang et al., 2009 ). (BS = 100%) + Huerteales + (Malvales + Brassicales; Within Sapindales, Nitrariaceae are sister to two clades (al- each with BS = 100%). These results are identical to recent though the sister-group relationship is not very well supported): April 2011] Soltis et al. — Angiosperm phylogeny 719

Fig. 2. Continued. 720 American Journal of Botany [Vol. 98

Fig. 2. Continued. April 2011] Soltis et al. — Angiosperm phylogeny 721

(1) Burseraceae + Anacardiaceae, and (2) a clade (also without COM clade— In Celastrales, Lepidobotryaceae are sister to a BS > 50%) of Sapindaceae sister to Meliaceae + (Simarou- broadly defi ned Celastraceae, following other recent focused baceae + Rutaceae). These relationships largely agree with analyses ( Simmons et al., 2001a , b ; Zhang and Simmons, 2006 ). other analyses (e.g., Gadek et al., 1996; Muellner et al., 2007); Within Oxalidales, Huaceae are sister to the remainder of the different relationships were reported by H. Wang et al. (2009), clade, which then forms two well-supported (BS = 100%) sub- albeit with poor sampling and low support. clades: (1) Connaraceae + Oxalidaceae, and (2) Brunelliaceae, In Crossosomatales, there are two well-supported (BS = 100%) Cephalotaceae, Cunoniaceae, and Elaeocarpaceae. Within the clades: Staphyleaceae + [Stachyuraceae + Crossosomataceae] second clade, Brunelliaceae are sister to Elaeocarpaceae + and Aphloiaceae + Strasburgeriaceae (including Ixerbaceae). (Cephalotaceae + Cunoniaceae). These relationships differ only In Myrtales, two major clades were recovered: one comprising slightly from previous analyses. In Zhang and Simmons (2006) , Onagraceae + Lythraceae and the other comprising the large results were well supported and identical to those here, but Ce- families Myrtaceae and Melastomataceae and associated phalotaceae were not sampled. In Wurdack and Davis (2009) a smaller families. These results are consistent with all previous weakly supported (BS = 59%) Cephalotaceae + Cunoniaceae studies of the order (e.g., Conti et al., 1997 ; Sytsma et al., 2004 ). were unresolved with Brunelliaceae and Elaeocarpaceae. In H. The placement of Combretaceae, however, is not strongly sup- Wang et al. (2009), Cunoniaceae were sister to (Cephalotaceae + ported, a result that is in agreement with previous studies. Elaeocarpaceae), albeit without BS support > 50%; the cur- rent study also has increased representation of families within Fabidae— Fabidae is well supported (BS = 99%), with Zy- Oxalidales over the H. Wang et al. (2009 ) analysis of Rosidae . gophyllales sister (albeit with low support, BS = 57%) to two Malpighiales are well represented in the current study, and major clades (each with BS = 100%): the nitrogen-fi xing clade relationships largely agree with recent analyses (e.g., Davis and (Cucurbitales, Fabales, Fagales, and Rosales) and a clade of Wurdack, 2004; Davis et al., 2001, 2005 ; Wurdack and Davis, Celastrales, Oxalidales (including Huaceae; see APG III), and 2009; Ruhfel et al., 2011). The strong agreement between our Malpighiales (the COM group; Endress and Matthews, 2006). results and those of Wurdack and Davis (2009) is not surprising These results agree with H. Wang et al. (2009 ). given that both studies share considerable data. We overview In the COM clade, Celastrales (BS = 100%) are sister to a the major features of this clade below. clade (BS = 59%) of Malpighiales (BS =100%) + Oxalidales We recovered a well-supported (BS = 99%) parietal-placen- (BS = 100%). This topology agrees with recent analyses (e.g., tation clade (sensu Wurdack and Davis, 2009). Within this Soltis et al., 2007 ; H. Wang et al., 2009 ; Wurdack and Davis, clade, Lacistemataceae are the well-supported sister to Salica- 2009 ). Alternative topologies in which Malpighiales and Oxal- caeae, and Goupiaceae are weakly supported (BS = 68%) as idales did not form a clade have been strongly rejected ( Wurdack sister to Violaceae. The placement of Goupiaceae was unre- and Davis, 2009 ). solved in Wurdack and Davis (2009). The clusioid clade sensu Wurdack and Davis (2009) was well supported (BS = 100%), Nitrogen-fi xing clade —Within the nitrogen-fi xing clade and within this clade Calophyllaceae were sister (BS = 74%) to (BS = 100%), Fabales, Fagales, Rosales, and Cucurbitales all Podostemaceae + Hypericaceae. These relationships are also in have BS = 100%. Fabales appear as sister with strong support agreement with Ruhfel et al. (2011). Balanopaceae are sister to (BS = 84%) to the remainder of the clade; Rosales are then a clade of Trigoniaceae + Dichapetalaceae, all of which is sister sister to a clade (BS = 78%) of Fagales + Cucurbitales. These to Chrysobalanaceae + Euphroniaceae. Other well-supported results are in agreement with recent studies (e.g., Soltis et al., (BS = 100%) sister-group relationships recovered here include: 2005 , 2007 ; H. Wang et al., 2009 ). Erythroxylaceae + Rhizophoraceae, Elatinaceae + Malpighi- In Rosales (Barbeyaceae and Dirachmaceae were not sam- aceae, Lophopyxidaceae + Putranjivaceae, and Phyllanthaceae + pled here), Rosaceae are sister to the remainder of the clade Picrodendraceae. The recently recognized (Wurdack and Davis, (BS = 100%), which forms two subclades: (1) a well-supported 2009 ) Euphorbiaceae segregate Peraceae are recovered here as clade (BS = 100%) of Ulmaceae sister to (Cannabaceae + (Ur- sister to Euphorbiaceae s.s. APG III (2009) has deferred recog- ticaceae + Moraceae)) and (2) a clade (BS = 73%) of Elae- nition of Peraceae pending additional support from other non- agnaceae + Rhamnaceae. These results agree with other recent mtDNA gene regions for placement of Raffl esiaceae within studies (e.g., Sytsma et al., 2002 ; H. Wang et al., 2009 ). Euphorbiaceae + Peraceae. Similar to recent fi ndings (i.e., Davis Relationships in Fagales also agree with broad (Soltis et al., et al., 2007 ; Wurdack and Davis, 2009 ), our preliminary 17- 2000 , 2007 ) as well as focused analyses ( Manos and Steele, gene analyses that sampled Raffl esiaceae (later excluded with 1997 ; Li et al., 2002 ). Nothofagaceae and Fagaceae are succes- most other parasites in our fi nal analyses, but available upon sive sisters to the remaining Fagales, which form two clades: (1) request) placed that holoparasitic lineage with this clade. De- Casuarinaceae + Betulaceae and (2) Myricaceae + Juglandaceae spite some advances in our understanding of Malpighiales here, (Ticodendraceae and Rhoipteleaceae were not sampled here). relationships of many constituent clades still remain unclear Our results for Cucurbitales are similar to recent analyses (see Wurdack and Davis, 2009 ). (with Corynocarpaceae and Tetrameleaceae not sampled here) (e.g., Zhang et al., 2006 ; Soltis et al., 2007 ). Anisophyllaceae Superasteridae— Superasteridae (BS = 87%) consists of and Coriariaceae are subsequent sisters to a clade (BS = 78%) Berberidopsidales, Santalales, Caryophyllales , Asteridae , and of (Datiscacaee + Begoniaceae) + Cucurbitaceae. possibly Dilleniaceae (see below). The 17-gene analysis placed In Fabales, Polygalaceae + Quillajaceae (BS support < 50%) Santalales as sister to a weakly supported (BS < 50%) clade are sister to a clade (BS = 62%) of Surianaceae + Fabaceae. The of (Berberidopsidales + Caryophyllales ) + Asteridae. The topologies for Fabales in recent analyses have varied greatly, sister grouping of Berberidopsidales + Caryophyllales received but most of these studies have involved only a few DNA re- BS = 75%. gions (see Wojciechowski et al., 2004 , Banks et al., 2008 ; Complete plastid genome data (Moore et al., 2010) and IR Bruneau et al., 2008 ; Bello et al., 2009 ). sequence data ( Moore et al., in press ) similarly recovered a 722 American Journal of Botany [Vol. 98 well-supported Superasteridae clade; hence, we formally name broadly circumscribed, are not monophyletic in our analyses, in this clade here (see Appendix 2). However, these three analyses agreement with previous studies (e.g., Applequist and Wallace, differ in the relationships suggested among members of Super- 2001; Cu é noud et al., 2002; Brockington et al., 2009; Nyffeler asteridae. Complete plastid genome sequence data place Santa- and Eggli, 2010 ), leading to the inclusion of Calyptrotheca in lales as sister to the remainder (BS = 100%); Berberidopsis is Didiereaceae and recognition of Montiaceae and Talinaceae (as then sister to a clade (BS = 88%) of Caryophyllales + Asteri- in APG III, 2009 ; Nyffeler and Eggli, 2010 ) and of Petiveri- dae . IR data suggest Santalales as sister to the well-supported aceae (see Judd et al., 2008 ; Nyffeler and Eggli, 2010 ). (BS = 98%) remaining taxa, in which Caryophyllales are sister to a clade (BS = 68%) of Berberidopsidales + Asteridae . Hence, Asteridae— Within Asteridae, Cornales are sister to the re- all three studies agree in the placement of Santalales as sister to maining taxa (BS = 97%) within which Ericales are sister to a the rest, but the precise placement of the enigmatic Berberidop- clade (BS = 100%) consisting of two subclades: Lamiidae sidales within Superasteridae remains unclear. (BS = 100%) + Campanulidae (BS = 100%). Recognition of Asteridae comprising these four groups was fi rst based on a Dilleniales— Dilleniaceae were recovered as sister to the re- molecular phylogenetic analysis ( Olmstead et al., 1992 ), and maining Superasteridae (BS = 97%). This position is much bet- the relationships among these groups are congruent with most ter supported than other recent placements of this enigmatic prior molecular studies (Olmstead et al., 2000; Albach et al., family (e.g., Moore et al., 2010 , in press ), but despite this high 2001 ; Bremer et al., 2002 ), although with stronger support. support, we remain cautious about its true placement because of differing placements in recent studies. Using complete plastid Cornales— Within Cornales (BS = 100%), a weakly sup- genome sequence data, Moore et al. (2010) placed Dilleniaceae ported clade (BS = 76%) of Cornus + Nyssa (Cornaceae s.l.) is as sister to Superrosidae (BS = 64%), but topology tests did not sister to a weakly supported (BS = 54%) clade of Curtisiaceae + reject alternative positions of Dilleniaceae as sister to Asteridae (Hydrangeaceae + Loasaceae). Only the sister pair of Hy- or to all remaining Pentapetalae . Analysis of IR sequences for drangeaceae + Loasaceae is well supported (BS = 100%), in 244 taxa placed Dilleniaceae sister to Superrosidae + Superas- agreement with other analyses (Fan and Xiang, 2003). Two teridae (BS = 79%) ( Moore et al., in press ). Because of these families, Grubbiaceae and Hydrostachyaceae, were not in- varied placements, Dilleniaceae are only tentatively placed here cluded here; the precise placement of the latter family within in Superasteridae based on their position in Fig. 1 . Additional Cornales has been particularly problematic ( Olmstead et al., research is required to place Dilleniaceae defi nitively. 2000 ; Albach et al., 2001 ; Fan and Xiang, 2003 ).

Santalales— Within Santalales (BS = 100%), the two sampled Ericales— The monophyly of Ericales is well supported (BS genera of “ Olacaceae ” , a family now regarded as nonmonophyl- = 100%), but as in other recent studies, resolution within the etic (see Mal é cot and Nickrent, 2008 ; Nickrent et al., 2010 ), are clade is more problematic. Our study retrieves a well-supported subsequent sisters to the well-supported (BS = 100%) remainder clade (BS = 100%) of Tetrameristaceae + (Marcgraviaceae + of the clade. Santalaceae and Opiliaceae are successive sisters to Balsaminaceae) (see Sch ö nenberger et al., 2010) as sister to the a clade (BS = 79%) of Schoepfi aceae + Loranthaceae. Represen- remaining Ericales, which are also well supported as monophyl- tation of Santalales is sparse, notably missing Erythropalaceae, etic (BS = 95%), but within which relationships are even more which may be sister to the rest of Santalales, and the holoparasitic uncertain, mirroring earlier analyses by Sch ö nenberger et al. Balanophoraceae, which have recently been recognized as part of (2005) based on 11 genes. Lecythidaceae are strongly supported this clade. Relationships among major clades of Santalales have (BS = 95%) as sister to the remainder, although in the 11-gene not been resolved with strong BS support in recent analyses, study with much greater taxon sampling ( Sch ö nenberger et al., and many segregate families have been newly recognized (e.g., 2005 ), Lecythidaceae were placed in a polytomy at this node. Mal é cot, 2002 ; Nickrent et al., 2010 ). Three other well-supported clades were recovered: (1) Po- lemoniaceae + Fouquieriaceae (BS = 92%), (2) a clade (BS = Caryophyllales— The topology resolved here is similar to 98%) of Actinidiaceae + Sarraceniaceae sister to a clade (BS = that of Brockington et al. (2009). Two major subclades were 100%) of Clethraceae + (Cyrillaceae + Ericaceae) (BS = 100%), recovered: core Caryophyllales (Caryophyllineae) and noncore and (3) a Primulaceae s.l. clade (BS = 100%) in which Maesa is Caryophyllales (Polygonineae) (clade names of Judd et al., sister to Clavija + Primulaceae s.s. (BS = 100%). As in the 11- 2008 ). Polygonineae in turn comprise two clades: (1) Plumbag- gene analysis (Sch ö nenberger et al., 2005), the relationships of inaceae + Polygonaceae and Tamaricaceae + Frankeniacae, and Diapensiaceae, Ebenaceae, Sapotaceae, Ternstroemiaceae, and (2) Droseraceae + Nepenthaceae sister to Drosophyllaceae + Theaceae are not well supported here. A recent 23-gene analysis (Ancistrocladaceae + Dioncophyllaceae). with approximately 90 taxa of Ericales resolves many of these Within Caryophyllineae, Rhabdodendraceae, Simmondsi- latter placements (K. Sytsma et al., unpublished manuscript). aceae, and Asteropeiaceae + Physenaceae are subsequent sis- ters to the rest of the clade. Within the latter, Caryophyllaceae + Lamiidae— Lamiidae are well supported (BS = 100%). (Amaranthaceae + Achatocarpaceae) are sister to the remain- Within this clade, Icacinaceae + Garryales, and Vahliaceae are der. In the remaining clade, Stegnospermataceae followed by subsequent sisters to the remaining lamiids, which comprise Limeaceae are sister to (1) Barbeuiaceae, Aizoaceae, and four major clades: Boraginaceae, Lamiales, Solanales, and Gisekiaceae as subsequent sisters to a clade of Phytolaccaceae Gentianales, among which there is no well-supported resolu- followed by Sarcobataceae as sisters to Petiveriaceae (as repre- tion. The position of Vahliaceae as immediate sister to the rest sented by Rivinia ) + Nyctaginaceae; and (2) Molluginaceae fol- of Lamiidae is consistent with the parsimony analysis of N. F. lowed by Montiaceae sister to two clades: Halophytaceae sister Refulio-Rodriguez and R. G. Olmstead (unpublished manu- to Basellaceae + Didiereaceae, and Talinaceae sister to Portu- script); however, in their ML and Bayesian analyses, Vahlia is lacaceae + Cactaceae. Portulacaceae and Phytolaccaceae, as sister to Solanales. April 2011] Soltis et al. — Angiosperm phylogeny 723

A more thorough analysis of the clade is provided by N. F. with strong support (BS = 100%). The several smaller Cam- Refulio-Rodriguez and R. G. Olmstead (unpublished manu- panulidae clades, including Escalloniales (= Escalloniaceae script) with a total of 129 species of Lamiidae, including all sensu Tank and Donoghue, 2010 ) and Paracryphiales (= recognized families (APG III, 2009). Their results differ mostly Paracryphiaceae sensu Tank and Donoghue, 2010 ), also receive in places where resolution is weak in the results presented here, strong support in the current study (BS = 91% and 99%, respec- such as the branching order among the early-diverging groups, tively). Although recovered in the best ML tree ( Figs. 1 and 2 ), Oncothecaceae, Icacinaceae, and Garryales, and among the bootstrap support for Bruniales (the Columelliaceae-Bruniaceae four major clades of Lamiidae , Lamiales, Solanales, Genti- clade sensu Tank and Donoghue, 2010 ) is < 50% in the total evi- anales, and Boraginaceae. dence analysis reported here. Relationships among the primary Campanulidae clades also Lamiales— Lamiales (BS = 100%) comprise Plocospermata- agree with previous analyses. We recovered the sister-group rela- ceae, Oleaceae, and Tetrachondraceae as well-supported subse- tionship of Aquifoliales and Apiidae (i.e., the remaining Cam- quent sisters (each with BS = 100%) to the rest of Lamiales. panulidae taxa; Cantino et al., 2007 ). Likewise, relationships Peltanthera (unassigned to family) plus (Gesneriaceae + Cal- among the six major lineages within Apiidae — Apiales, Aster- ceolariaceae) form a clade (BS = 97%) sister to the remaining ales, Bruniales, Dipsacales, Escalloniaceae, and Paracryphi- Lamiales, which form a strongly supported core (BS = 100%). aceae — were identical to earlier results (e.g., Zhang et al., 2003 ; Within this core, few relationships have BS values > 50% in our Winkworth et al., 2008 ; Tank and Donoghue, 2010 ). To facilitate analyses, indicating the diffi culties in resolving relationships in the study of evolutionary patterns within Apiidae , Tank and this clade. Relationships with support > 70% include Verben- Donoghue (2010) named several clades that they recovered with aceae + Thomandersiaceae (BS = 75%), and (((Phrymaceae + strong statistical support, including Dipsapiidae (the Apiales- Orobanchaceae (BS = 95%)) + Paulowniaceae (BS = 99%)) + Dipsacales-Paracryphiaceae clade) and Dipsidae (the Dipsacales- Lamiaceae (BS = 74%)). These results are consistent with those Paracryphiaceae clade). Both of these clades were recovered in obtained by Refulio-Rodriguez and Olmstead (unpublished the total evidence analysis, albeit with < 50% bootstrap support. manuscript), except that Orobanchaceae and Paulownia are well-supported sister groups, followed by Phrymaceae in their Aquifoliales— Aquifoliales were resolved as the sister group results. However, most of these families are represented here by to Apiidae. Within Aquifoliales, there are two major lineages: only a single species. Cardiopteridaceae + Stemonuraceae and a clade of Aquifoli- aceae + (Phyllonomaceae + Helwingiaceae). As did Tank and Solanales— The topology for Solanales agrees with other re- Donoghue (2010) , we found strong support for Phyllonoma- cent analyses, although most have not had complete family Helwingia as sister to Ilex ( Fig. 2 ). Although Phyllonoma and coverage of the order (e.g., Soltis et al., 2000 ). Complete family Helwingia share several conspicuous morphological apomor- coverage was obtained in Bremer et al. (2002), although the phies, including epiphyllous infl orescences, most earlier studies current study provides much stronger internal support because (based on far less data) placed Helwingia directly with Ilex to of the large number of genes included and added sampling the exclusion of Phyllonoma (e.g., Morgan and Soltis, 1993; within Solanaceae and Convolvulaceae. We found Montini- Soltis and Soltis, 1997 ; Olmstead et al., 2000 ). aceae are sister to Spenocleaceae + Hydroleaceae; this well- supported clade (BS = 98%) is in turn sister to Convolvulaceae + Escalloniaceae— Escalloniaceae (= Escalloniales sensu APG Solanaceae (BS = 100%). III, 2009 ) were recovered with strong support ( Figs. 1 and 2 ; BS = 91%); however, their relationship to the other lineages of Cam- Gentianales— Rubiaceae are sister to the remainder of the panulidae remain uncertain. In addition, relationships among the clade, in agreement with other studies (e.g., Backlund et al., genera are not well supported aside from a core Escalloniaceae 2000; Frasier, 2009). Relationships among the remaining fami- clade composed of Escallonia , Forgesia , and Valdivia ( Fig. 2 ). lies have been unclear in most previous studies (e.g., Olmstead In our mtDNA analyses, which included Escallonia , Eremosyne , et al., 2000 ; Soltis et al., 2000 ; Bremer et al., 2002 ), but here and Polyosma , none of the genes recovered a monophyletic Es- Loganiaceae are weakly supported as sister to Gelsemiaceae + calloniaceae. However, there is little support for relationships (Apocynaceae + Gentianaceae), which is congruent with a among any of the unresolved lineages in the mtDNA analyses study of four plastid DNA regions ( Frasier, 2009 ). Although the (Appendix S6), with one exception: when mtDNA data are in- current study does reveal areas of strong support, sampling in cluded for Polyosma , it is placed with strong support within the families of this order is low. Paracryphiaceae (= Paracryphiales sensu APG III, 2009). With- out the mtDNA data, Polyosma is placed with BS = 91% in Es- Campanulidae— A Campanulidae clade was recovered with calloniaceae, as expected based on previous analyses (e.g., strong support (BS = 100%), and relationships within largely Lundberg, 2001 ; Winkworth et al., 2008 ; Tank and Donoghue, agree with recent analyses, especially Tank and Donoghue (2010 ; 2010 ) (compare Fig. 2 and Appendix S3; see below, Comparison also see Lundberg, 2001; Winkworth et al., 2008). This is not of mtDNA, rDNA, and plastid topologies ). surprising given the very large overlap in data between the cur- rent study and that of Tank and Donoghue (2010) , the major dif- Asterales— Each of the major lineages of Asterales was re- ference being the addition of mtDNA and nuclear rDNA data in covered with strong support, and, for the most part, relation- the current study. Tank and Donoghue (2010) used their phylo- ships among them are also well supported (Fig. 2). One genetic results to discuss fl oral evolution in Campanulidae , and exception is at the base of the Asterales, where the successive analyses of biogeographic patterns, especially in relation to past sister-group relationship of Pentaphragmataceae and the Rous- continental connections, are underway. saceae + Campanulaceae clade to the remainder of Asterales Each of the four major lineages of Campanulidae — Apiales, received < 50% bootstrap support (Fig. 2 ). In earlier studies, these Aquifoliales, Asterales, and Dipsacales— was recovered here relationships were reversed, with the Roussaceae-Campanulaceae 724 American Journal of Botany [Vol. 98 clade sister to the rest of Asterales ( Lundberg and Bremer, Comparison of mtDNA, rDNA, and plastid topologies — In 2003 ; Winkworth et al., 2008 ; Tank and Donoghue, 2010 ). In the present study, there is some confl ict among trees derived our analyses of the plastid partition alone, we recovered the from mtDNA, rDNA, and plastid data. However, resolution and same relationships as the earlier studies with strong support support of the rDNA topology is so low that there are no in- (Appendix S5). By contrast, the mtDNA partition provides strong stances of mutual incongruence with BS greater than 75% in- support for the alternative resolution of these relationships (Ap- volving it and either of the other partitions. pendix S6). When the mtDNA gene trees are examined sepa- One major case of incongruence involves relationships among rately, atp1 and matR recovered the mtDNA topology (BS < 50%), major lineages of Rosidae . The analysis of mitochondrial genes while rps3 resolved the plastid topology (BS values < 50%) (Ap- shows that the weakly supported COM clade is sister to the pendices S17 – S20). strongly supported core members of Malvidae (Sapindales, Mal- vales, Brassicales, Huerteales, Picramniaceae) with 92% BS sup- Bruniales— As noted above, Bruniales were recovered with port. This result parallels results obtained in a recent mtDNA strong support by Tank and Donoghue (2010), but were only analysis of angiosperms ( Qiu et al., 2010) . An earlier analysis of weakly supported (< 50% bootstrap) in our combined analysis. matR also recovered the same relationship, albeit with lower boot- This case is discussed further below (see Comparison of mtDNA, strap support ( Zhu et al., 2007 ). In all previous large-scale analy- rDNA, and plastid topologies ). This clade is of biogeographic ses of angiosperms or eudicots, the COM clade is sister to the interest in that Columelliaceae are restricted to South America, nitrogen-fi xing clade ( Chase et al., 1993 ; Savolainen et al., 2000a , while Bruniaceae are a primarily South African lineage; this b ; Soltis et al., 2000; Hilu et al., 2003; Burleigh et al., 2009), oc- appears to be an ancient vicariance event in Campanulidae , and casionally with high support (e.g., BS = 89% in Burleigh et al., the detailed historical biogeography of this clade is the subject 2009 ). These analyses were based either entirely on plastid genes of current research. alone or in combination with nuclear rDNA (18S and 26S rDNA). Thus far, only one large-scale analysis of angiosperms has been Dipsapiidae— Tank and Donoghue (2010) named this clade to performed on nuclear gene data, 18S rDNA (Soltis et al., 1997). In emphasize the close relationship between Apiales and Dipsacales that study, clades corresponding to Fabidae and Malvidae were (along with Paracryphiaceae, = Paracryphiales sensu APG III) — not recovered, but this has been a general problem with 18S and a result that appeared in several earlier studies (e.g., Bremer et al., 26S rDNA data — limited resolution and low support due to low 2002; Winkworth et al., 2008) and is strongly supported in their phylogenetic signal (see Soltis et al., 1997 ). plastid analyses ( Tank and Donoghue, 2010 ). Surprisingly, this One line of evidence that lends some support to the possible clade is not recovered with > 50% bootstrap support in our total sister relationship between the COM clade and Malvidae sug- evidence analyses, although it does appear in the ML BS tree gested by mtDNA data comes from a broad survey of fl oral ( Figs. 1 and 2 ). Within Dipsapiidae , Paracryphiaceae are resolved structural characters ( Endress and Matthews, 2006 ). Twenty- as sister to Dipsacales (= Dipsidae sensu Tank and Donoghue, two COM families and 18 families of Malvidae share a type 2010 ), albeit with < 50% bootstrap support. As with Dipsapiidae , of ovule with a thicker inner integument than the outer one, a Dipsidae were strongly supported in the analyses of both situation that is otherwise very rare in eudicots (with only one Wink worth et al. (2008) and Tank and Donoghue (2010) . other occurrence, in Trochodendrales). That survey did not fi nd Relationships within Dipsacales largely agree with earlier any morphological synapomorphy supporting the monophyly of studies (e.g., Bell et al., 2001 ; Donoghue et al., 2001 , 2003 ; Fabidae , but this is not surprising in that morphological synapo- Bell, 2004 ; Winkworth et al., 2008 ; Tank and Donoghue, 2010 ), morphies remain elusive for many major clades of angiosperms. as do relationships within Apiales (e.g., K å rehed, 2001, 2003 ; Likewise, within Campanulidae there are several instances Chandler and Plunkett, 2004; Winkworth et al., 2008; Tank and where mtDNA and plastid analyses are in confl ict. The most Donoghue, 2010), with only minor differences in areas with obvious example of confl ict among data sets involved the place- little to no bootstrap support (e.g., within Araliaceae). ment of Polyosma . With mtDNA sequence data included, Poly- osma (Escalloniaceae) is placed sister to Quintinia in the Possible taxon sampling issues— Several relationships do not Paracryphiales (Appendix S2), but when mtDNA data for Poly- agree with more detailed phylogenetic studies focused on partic- osma are removed, it is placed in Escalloniaceae. Both alterna- ular clades (see also Tank and Donoghue, 2010 ). For example, tive placements received high BS support, and nuclear rDNA within Dipsacaceae (Dipsacales), Pterocephalodes is sister to data do not provide suffi cient resolution of the placement of this Dipsacus, and Scabiosa is their subsequent sister. The sister taxon. These results suggest either a biological phenomenon group of Pterocephalodes and Dipsacus was very strongly sup- (e.g., HGT of mtDNA from some member of Paracryphiaceae ported in the analyses of Tank and Donoghue (2010), but is less to the Polyosma lineage) or human error (e.g., incorrect label- strongly supported here (BS = 54%). This result is incongruent ing of material). All of the plastid sequences for Polyosma came with the much more densely sampled analyses of Carlson et al. from two accessions — one used in Bremer et al. (2002) and one (2009), where Pterocephalodes emerged as sister to all other a silica-gel collection obtained from MO (Tank and Donoghue, Dipsacaceae. In contrast, while Tank and Donoghue (2010) 2010 ). The individual plastid gene trees all agree with the place- found a well-supported sister-group relationship between Echin- ment of Polyosma in Escalloniaceae. The atp1 sequence for ops and Gerbera within Asteraceae, here we recovered the suc- Polyosma was obtained from our MO accession, while the other cessive sister-group relationship of Gerbera and Echinops to the three mtDNA genes came from a separate accession (Kew remainder of Asteraceae that appears in more detailed studies of M285-TL265). Because sequences from the MO accession used Asteraceae (e.g., Panero and Funk, 2008 ). In general, these ex- by Tank and Donoghue (2010) in their plastid analyses agree amples highlight that although statistical support may be high for with previous studies based on a different accession (e.g., particular relationships, taxon and gene sampling can both have Bremer et al., 2002), it is possible that the M285-TL265 acces- major effects on the outcome, especially when the diversity of sion was misidentifi ed, or tissue or tubes were mislabeled. the group in question has been grossly undersampled. Signifi cantly, sequencing of the mitochondrial genes surveyed April 2011] Soltis et al. — Angiosperm phylogeny 725 here in a second accession of Polyosma , as well as a second lies of fl owering plants: APG III. Botanical Journal of the Linnean sample of Quintinia (the sister of Polyosma when mtDNA are Society 161: 105 – 121. included), yielded nearly identical sequences and the same to- Applequist , W. L. , and R. S. Wallace . 2001 . Phylogeny of the por- pologies. Thus, available data favor HGT as the more likely tulacaceous cohort based on ndhF sequence data. Systematic Botany cause of incongruence involving Polyosma . 26 : 406 – 419 . Backlund , M. , B. Oxelman , and B. Bremer. 2000 . Phylogenetic Further evidence of confl ict between the mtDNA and plastid relationships within the Gentianales based on ndhF and rbcL se- partitions in Campanulidae comes from the lack of bootstrap quences, with particular reference to the Loganiaceae. American Journal support in the combined analyses for several clades that were of Botany 87 : 1029 – 1043 . well supported in our analyses of the plastid data alone. In anal- Banks , H. , B. B. Klitgaard , F. Claxton , F. Forest , and P. R. Crane . yses of the plastid partition (Appendix S5), we recovered a 2008 . Pollen morphology of the family Polygalaceae (Fabales). Bruniales clade with 79% BS support, in agreement with Tank Botanical Journal of the Linnean Society 156 : 253 – 289 . and Donoghue (2010) . However, as noted above, in the com- Barkman , T. J. , J. R. McNeal , S. H. Lim , G. Coat , H. B. Croom , bined analysis this relationship received < 50% bootstrap sup- N. D. Young , and C. W. Depamphilis . 2007 . Mitochondrial DNA port. These results suggest that there is enough confl ict between suggests at least 11 origins of parasitism in angiosperms and reveals the mtDNA partition and the plastid partition to collapse this genomic chimerism in parasitic plants. BMC Evolutionary Biology 7 : 248 . relationship. We observed several other such cases, including Bayer , C. , M. F. Fay , A. de Bruijn , V. Savolainen , C. M. Morton , K. the Dipsapiidae and Dipsidae clades and relationships at the Kubitzki , and M. W. Chase . 1999 . Support for an expanded fam- base of Asterales, as noted above. ily concept of Malvaceae within a recircumscribed order Malvales: A combined analysis of plastid atpB and rbcL sequences. Botanical Impact of taxon density— One concern with the complete Journal of the Linnean Society 129 : 267 – 303 . plastid genome analysis of angiosperms by Moore et al. (2010) Beaman , R. S. , and N. Cellinese . 2010 . TOLKIN: The Tree of Life is that only 86 taxa were employed, raising questions regarding Knowledge and Information Network. Website http://www.tolkin. the impact of taxon density on the topology. To test this, we org/ . trimmed the 17-gene data set from 640 taxa to the 86 taxa in the Bell , C. D. 2004 . Preliminary phylogeny of Valerianaceae (Dipsacales) analysis by Moore et al. (2010) . inferred from nuclear and chloroplast DNA sequence data. Molecular Phylogenetics and Evolution 31 : 340 – 350 . The topologies of the 17-gene 640-taxon and reduced 86- Bell , C. D. , E. J. Edwards , S. T. Kim , and M. J. Donoghue . 2001 . taxon (Appendix S21) data sets are very similar, but with a few Dipsacales phylogeny based on chloroplast DNA sequences. Harvard differences. Topological similarities include strong support for: Papers in Botany 6 : 481 – 499 . Chloranthaceae + Magnoliidae (BS = 94%); Monocotyledon- Bell , C. D. , D. E. Soltis , and P. Soltis . 2010 . The age and diversifi - eae + (Ceratophyllaceae + Eudicotyledoneae ) (BS = 97%); and cation of angiosperms re-revisited. American Journal of Botany 97 : Trochodendron followed by Buxus as subsequent sisters to 1296 – 1303 . Gunneridae (BS = 100%; 95%). However, there are notewor- Bello , M. A. , A. Bruneau , F. Forest , and J. A. Hawkins . 2009 . thy differences between the trees. First, in the reduced tree, Elusive relationships within order Fabales: Phylogenetic analyses us- Amborella is sister to Nymphaeales (BS = 97%), rather than ing matK and rbcL sequence data. Systematic Botany 34 : 102 – 114 . Amborella followed by Nymphaeales as sister to all other ex- Bergthorsson , U. , K. L. Adams , B. Thomason , and J. D. Palmer . tant angiosperms. Second, Nelumbonaceae and Meliosma are 2003 . Widespread horizontal transfer of mitochondrial genes in fl owering plants. Nature 424 : 197 – 201 . subsequent sisters to other angiosperms in the reduced tree, Bowe , L. M. , and C. W. dePamphilis . 1996 . Effects of RNA editing and rather than a clade, as in the larger tree. Third, Superrosidae are gene processing on phylogenetic reconstruction. Molecular Biology not recovered in the reduced tree; instead, a modifi ed Superas- and Evolution 13 : 1159 – 1166 . teridae (BS = 50%) includes Vitaceae (a member of Superrosi- Bremer , B. , K. Bremer , N. Heidari , P. Erixon , R. G. Olmstead , dae in the complete analysis). Fourth, Dilleniaceae are sister to A. A. Anderberg , M. K ä llersj ö , and E. Barkhordarian . 2002 . Caryophyllales (BS < 50%), rather than sister to all Superas- Phylogenetics of asterids based on 3 coding and 3 non-coding teridae . This placement of Dilleniaceae as sister to Caryophyl- chloroplast DNA markers and the utility of non-coding DNA at lales was also seen in the three-gene tree (e.g., Soltis et al., higher taxonomic levels. Molecular Phylogenetics and Evolution 24 : 2000 ), but not in the larger tree here. All of these placements 274 – 301 . not only differ from the 17-gene, 640-taxon analysis, but also Brockington , S. F. , R. Alexandre , J. Ramdial , M. J. Moore , S. Crawley , A. Dhingra , K. Hilu , et al . 2010 . Phylogeny of the Caryophyllales from the complete plastid genome topology and from an analy- sensu lato: Revisiting hypotheses on pollination biology and perianth sis of the IR for 244 taxa ( Moore et al., in press ). These results differentiation in the Core Caryophyllales. International Journal of further confi rm that reduced taxon density can have a major Plant Sciences 171 : 185 – 198 . impact on topology, even when numerous genes are employed. Bruneau , A. , M. Mercure , G. P. Lewis , and P. S. Herendeen . 2008 . Furthermore, some of these differences could also refl ect con- Phylogenetic patterns and diversifi cation in the caesalpinioid legumes. fl ict between plastid and mtDNA data. Botany 86 : 697 – 718 . Burleigh , J. G. , K. W. Hilu , and D. E. Soltis . 2009 . Inferring phy- logenies with incomplete data sets: a 5-gene, 567-taxon analysis of LITERATURE CITED angiosperms. BMC Evolutionary Biology 9 : 61 . Cantino , P. D. , J. A. Doyle , S. W. Graham , W. S. Judd , R. G. Olmstead , Albach , D. C. , D. E. Soltis , M. W. Chase , and D. E. Soltis . 2001 . D. E. Soltis , P. S. Soltis , and M. J. Donoghue . 2007 . Towards a Phylogenetic placement of the enigmatic angiosperm Hydrostachys. phylogenetic nomenclature of Tracheophyta. Taxon 56 : 822 – 846 . Taxon 50 : 781 – 805 . Carlson , S. E. , V. Mayer , and M. J. Donoghue . 2009 . Phylogenetic re- Alverson , W. S. , B. A. Whitlock , R. Nyffeler , C. Bayer , and D. A. lationships, taxonomy, and morphological evolution in Dipsacaceae Baum. 1999 . Phylogeny of the core Malvales: Evidence from ndhF (Dipsacales) inferred by DNA sequence data. Taxon 58 : 1075 – 1091 . sequence data. American Journal of Botany 86 : 1474 – 1486 . Castresana , J. 2000 . Selection of conserved blocks from multiple align- APG III [Angiosperm Phylogeny Group III]. 2009 . An update of the ments for their use in phylogenetic analysis. Molecular Biology and Angiosperm Phylogeny Group classifi cation for the orders and fami- Evolution 17 : 540 – 552 . 726 American Journal of Botany [Vol. 98

Chandler , G. T. , and G. M. Plunkett . 2004 . Evolution in Apiales: nu- University of Washington, Seattle, Washington, USA. Website http:// clear and chloroplast markers together in (almost) perfect harmony. www.phylip.com/. Botanical Journal of the Linnean Society 144 : 123 – 147 . Frasier , C. L. 2009 . Evolution and systematics of the angiosperm order Chase , M. , D. E. Soltis , D. Morgan , D. Olmsted , M. Duvall , R. Price , Gentianales with an in-depth focus on Loganiaceae and its species- H. Hills , et al . 1993 . Phylogenetic relationships among seed plants rich and toxic genus Strychnos. PhD dissertation, Rutgers University, based on rbcL sequence data. Annals of the Missouri Botanical New Brunswick, New Jersey, USA. Garden 80 : 528 – 580 . Funk , V. A. , A. Susanne , T. F. Stuessy , and R. J. Bayer , eds. 2009 . Chase , M. W. , M. F. Fay , D. Devey , O. Maurin , N. R ø nsted , J. Davies , Systematics, evolution, and biogeography of Compositae. International Y. Pillon , et al . 2006 . Multigene analyses of monocot relationships: Association for Plant Taxonomy, Institute of Botany, University of a summary. In J. T. Columbus, E. A. Friar, J. M. Porter, L. M. Prince, Vienna. Vienna, Austria. and M. G. Simpson [eds.], Monocots: Comparative biology and Gadek , P. A. , E. S. Fernando , C. J. Quinn , S. B. Hoot , T. Terrazas , evolution. Excluding Poales. Rancho Santa Ana Botanical Garden, M. C. Sheahan , and M. W. Chase . 1996 . Sapindales: Molecular Claremont, California, USA. Aliso 22: 63 – 75. delimitation and infraordinal groups. American Journal of Botany Conti , E. , A. Litt , P. G. Wilson , S. A. Graham , B. G. Briggs , L. A. 83 : 802 – 811 . S. Johnson , and K. J. Sytsma . 1997 . Interfamilial relationships in Givnish , T. J. , J. H. Leebens-Mack , M. Ames Sevillano , J. R. Myrtales: Molecular phylogeny and patterns of morphological evolu- McNeal , P. R. Steele , J. I. Davis , C. Ané , et al. 2010 . Assembling tion. Systematic Botany 22 : 629 – 647 . the tree of the monocotyledons: Plastome sequence phylogeny and Cuenoud , P. , V. Savolainen , L. W. Chatrou , M. Powell , R. J. Grayer , evolution of Poales. Annals of the Missouri Botanical Garden and M. W. Chase . 2002 . Molecular phylogenetics of Caryophyllales 97: 584– 616. based on nuclear 18S rDNA and plastid rbcL, atpB , and matK DNA Golubchik , T. , M. J. Wise , S. Easteal , and L. S. Jermiin . 2007 . Mind the sequences. American Journal of Botany 89 : 132 – 144 . gaps: Evidence of bias in estimates of multiple sequence alignments. Davies , T. J. , T. G. Barraclough , M. W. Chase , P. S. Soltis , D. E. Soltis , Molecular Biology and Evolution 24 : 2433 – 2442 . and V. Savolainen. 2004 . A supertree of the angiosperms. Proceed- Graham , S. W. , J. M. Zgurski , M. A. McPherson , D. M. Cherniawsky , ings of the National Academy of Sciences, USA 101 : 1904 – 1909 . J. M. Saarela , E. F. C. Horne , S. Y. Smith , et al. 2006 . Robust Davis , C. C. , W. R. Anderson , and M. J. Donoghue . 2001 . Phylogeny of inference of monocot deep phylogeny using an expanded multigene Malpighiaceae: Evidence from chloroplast ndhF and trnL-F nucle- plastid data set. In J. T. Columbus, E. A. Friar, J. M. Porter, L. M. otide sequences. American Journal of Botany 88 : 1830 – 1846 . Prince, and M. G. Simpson [eds.], Monocots: Comparative biol- Davis , C. C. , M. Latvis , D. L. Nickrent , K. J. Wurdack , and D. A. Baum . ogy and evolution. Excluding Poales. Rancho Santa Ana Botanical 2007 . Floral gigantism in Raffl esiaceae. Science 315 : 1812 . Garden, Claremont, California, USA. Aliso 22: 3– 21. Davis , C. C. , C. O. Webb , K. J. Wurdack , C. A. Jaramillo , and M. J. Hall , J. C. , H. H. Iltis , and K. J. Sytsma. 2004 . Molecular phlyoge- Donoghue . 2005 . Explosive radiation of Malpighiales supports netics of core Brassicales, placement of orphan genera Emblingia, a Mid-Cretaceous origin of modern tropical rain forests. American Forchammeria, Tirania, and character evolution. Systematic Botany Naturalist 165 : E36 – E65 . 29 : 654 – 669 . Davis , C. C. , and K. J. Wurdack . 2004 . Host-to-parasite gene trans- Hilu , K. W. , T. Borsch , K. Muller , D. E. Soltis , P. S. Soltis , V. fer in fl owering plants: phylogenetic evidence from Malpighiales. Savolainen , M. W. Chase , et al . 2003 . Angiosperm phylogeny Science 305 : 676 – 678 . based on matK sequence information. American Journal of Botany 9 0 : Donoghue , M. J. , C. D. Bell , and R. C. Winkworth . 2003 . The evo- 1758 – 1776 . lution of reproductive characters in Dipsacales. International Journal Jansen , R. K. , Z. Cai , L. A. Raubeson , H. Daniell , C. W. dePamphilis , of Plant Sciences 164 ( 5 supplement ): S453 – S464 . J. Leebens-Mack , K. F. Mü ller , et al . 2007 . Analysis of 81 genes Donoghue , M. J. , T. Eriksson , P. A. Reeves , and R. G. Olmstead . 2001 . from 64 plastid genomes resolves relationships in angiosperms and Phylogeny and phylogenetic taxonomy of Dipsacales, with special identifi es genome-scale evolutionary patterns. Proceedings of the reference to Sinadoxa and Tetradoxa (Adoxaceae). Harvard Papers National Academy of Sciences, USA 104 : 19369 – 19374 . in Botany 6 : 459 – 479 . Jian , S. , P. S. Soltis , M. Gitzendanner , M. Moore , R. Li , T. Hendry , Doyle , J. J. , and J. L. Doyle . 1987 . A rapid DNA isolation procedure for Y. Qiu , et al . 2008 . Resolving an ancient, rapid radiation in small quantities of fresh leaf tissue. Phytochem. Bull. 19 : 11 – 15 . Saxifragales. Systematic Biology 57 : 38 – 57 . Doyle , J. A. , H. Eklund , and P. S. Herendeen. 2003 . Floral evo- Jiao , Y. , N. Wickett , S. Ayyampalayam , A. Chanderbali , L. Landherr , lution in Chloranthaceae: Implications of a morphological phy- P. E. Ralph , P. S. Soltis , et al. In press. Phylogenomic analysis re- logenetic analysis. International Journal of Plant Sciences 164 : veals ancient genome duplications in seed plant and angiosperm his- S365 – S382 . tory. Nature . Eklund , H. , J. A. Doyle , and P. S. Herendeen . 2004 . Morphological Judd , W. S. , C. S. Campbell , E. A. Kellogg , P. F. Stevens , and M. J. phylogenetic analysis of living and fossil Chloranthaceae. International Donoghue. 2008 . Plant systematics: A phylogenetic approach, 3rd Journal of Plant Sciences 165 : 107 – 151 . ed. Sinauer, Sunderland, Massachusetts, USA. Endress , P. K. , and M. L. Matthews . 2006 . First steps towards a Kå rehed , J. 2001 . Multiple origin of the tropical forest tree family fl oral structural characterization of the major rosid subclades. Plant Icacinaceae. American Journal of Botany 88 : 2259 – 2274 . Systematics and Evolution 260 : 223 – 251 . Kå rehed , J. 2003 . The family Pennantiaceae and its relationships to Evans , R. C. , L. A. Alice , C. S. Campbell , T. A. Dickinson , and E. A. Apiales. Botanical Journal of the Linnean Society 141 : 1 – 24 . Kellogg . 2000 . The granule-bound starch synthase (GBSSI) gene Katoh , K. , and H. Toh. 2008 . Recent developments in the MAFFT mul- in the Rosaceae: multiple loci and phylogenetic utility. Molecular tiple sequence alignment program. Briefi ngs in Bioinformatics 9 : Phylogenetics and Evolution 17 : 388 – 400 . 286 – 298 . Fan , C. , and Q.-Y. Xiang . 2003 . Phylogenetic analyses of Cornales Kim , S. , D. E. Soltis , P. S. Soltis , M. Zanis , and Y. Suh . 2004 . Phylo- based on 26S rRNA and combined 26S rRNA- matK -rbcL sequence genetic relationships among early-diverging eudicots: Evidence from data. American Journal of Botany 90 : 1357 – 1372 . atpB, rbcL , 18S rDNA, and 26S rDNA sequences. Molecular Felsenstein , J. 1973 . Maximum likelihood and minimum-steps methods Phylogenetics and Evolution 31 : 16 – 30 . for estimating evolutionary trees from data on discrete characters. Kluge , A. G. 1989 . A concern for evidence and a phylogenetic hypoth- Systematic Zoology 22 : 240 – 249 . esis for relationships among Epicrates (Boidae, Serpentes). Systematic Felsenstein , J. 1985 . Confi dence limits on phylogenies: An approach Zoology 38 : 7 – 25 . using the bootstrap. Evolution 39 : 783 – 791 . Lecointre , G. , and P. Deleporte . 2005 . Total evidence requires exclu- Felsenstein , J. 2005 . PHYLIP (Phylogeny Inference Package) version sion of phylogenetically misleading data. Zoologica Scripta 3 4 : 3.6. Distributed by the author. Department of Genome Sciences, 101 – 117 . April 2011] Soltis et al. — Angiosperm phylogeny 727

Leebens-Mack , J. , L. A. Raubeson , L. Cui , J. V. Kuehl , M. H. Fourcade , Qiu , Y.-L. , O. Dombrovska , J. H. Lee , L. B. Li , B. A. Whitlock , F. T. W. Chumley , J. L. Boore , et al . 2005 . Identifying the basal an- Bernasconi-Quadroni , J. S. Rest , et al . 2005 . Phylogenetic anal- giosperm node in chloroplast genome phylogenies: Sampling one ’ s yses of basal angiosperms based on nine plastid, mitochondrial, and way out of the Felsenstein zone. Molecular Biology and Evolution 2 2 : nuclear genes. International Journal of Plant Sciences 166 : 815 – 842 . 1948 – 1963 . Qiu , Y.-L. , J. Lee , F. Bernasconi-Quadroni , D. E. Soltis , P. S. Soltis , Li , R.-Q. , Z.-D. Chen , Y.-P. Hong , and A.-M. Lu . 2002 . Phylogenetic M. J. Zanis , E. A. Zimmer , et al . 1999 . The earliest angiosperms: relationships of the “ higher ” hamamelids based on chloroplast trnL-F Evidence from mitochondrial, plastid and nuclear genes. Nature 402 : sequences. Acta Botanica Sinica 44 : 1462 – 1468 . 404 – 407 . Lundberg , J. 2001 . Phylogenetic studies in the euasterids II with par- Qiu , Y.-L. , L. Li , B. Wang , J.-Y. Xue , T. A. Hendry , R.-Q. Li , J. W. ticular reference to Asterales and Escalloniaceae. Ph.D. dissertation, Brown , et al. 2010 . Angiosperm phylogeny inferred from sequences Uppsala University, Uppsala, Sweden. of four mitochondrial genes. Journal of Systematics and Evolution Lundberg , J. , and K. Bremer . 2003 . A phylogenetic study of the order 48: 391 – 425. Asterales using one morphological and three molecular data sets. Renner , S. S. 1999 . Circumscription and phylogeny of Laurales: Evidence International Journal of Plant Sciences 164 : 553 – 578 . from molecular and morphological data. American Journal of Botany Malé cot , V. 2002 . Histoire, classifi cation et phylog é nie des Olacaceae 86 : 1301 – 1315 . Brown (Santalales). Th è se de Doctorat de l ’ Universit é Paris 6, France. Rodman , J. E. , K. G. Karol , R. A. Price , and K. J. Sytsma . 1996 . Mal é cot , V. , and D. L. Nickrent . 2008 . Molecular phylogenetic rela- Molecules, morphology, and Dahlgren ’ s expanded order Capparales. tionships of Olacaceae and related Santalales. Systematic Botany 33 : Systematic Botany 21 : 289 – 307 . 97 – 106 . Rodman , J. E. , P. S. Soltis , D. E. Soltis , K. J. Sytsma , and K. G. Karol . Manos , P. S. , and K. P. Steele . 1997 . Phylogenetic analysis of “ higher ” 1998 . Parallel evolution of glucosinolate biosynthesis inferred from Hamamelididae based on plastid sequence data. American Journal of congruent nuclear and plastid gene phylogenies. American Journal of Botany 84 : 1407 – 1419 . Botany 85 : 997 – 1007 . Moore , M. J. , C. D. Bell , P. S. Soltis , and D. E. Soltis . 2007 . Using Rudall , P. J. , D. D. Sokoloff , M. V. Remizowa , J. G. Conran , plastid genome-scale data to resolve enigmatic relationships among J. I. Davis , T. D. Macfarlane , and D. W. Stevenson . 2007 . basal angiosperms. Proceedings of the National Academy of Sciences, Morphology of Hydatellaceae, an anomalous aquatic family recently USA 104 : 19363 – 19368 . recognized as an early-divergent angiosperm lineage. American Moore , M. J. , N. Hassan , M. A. Gitzendanner , R. A. Bruenn , M. Croley , Journal of Botany 94 : 1073 – 1092 . A. Vandeventer , J. W. Horn , et al . In press . Phylogenetic analysis Ruhfel , B. R. , V. Bittrich , C. P. Bove , M. H. G. Gustafsson , C. T. of the plastid inverted repeat for 244 species: Insights into deeper- Philbrick , R. Rutihauser , Z. Xi , and C. C. Davis. 2011 . Phylogeny of level angiosperm relationships from a long, slowly evolving sequence the clusioid clade (Malpighiales): evidence from the plastid and mito- region. International Journal of Plant Sciences . chondrial genomes. American Journal of Botany 98: 306 – 325. Moore, M. J. , P. S. Soltis , C. D. Bell , J. G. Burleigh , and D. E. Soltis . Saarela , J. M. , H. S. Rai , J. A. Doyle , P. K. Endress , S. Mathews , A. D. 2010 . Phylogenetic analysis of 83 plastid genes further resolves the Marchant , B. G. Briggs , and S. W. Graham . 2007 . Hydatellaceae early diversifi cation of eudicots. Proceedings of the National Academy identifi ed as a new branch near the base of the angiosperm phyloge- of Sciences, USA 107 : 4623 – 4628 . netic tree. Nature 446 : 312 – 315 . Morgan , D. R. , and D. E. Soltis . 1993 . Phylogenetic relationships Sanderson , M. J. , M. M. McMahon , and M. Steel . 2010 . Phylogenomics among members of Saxifragaceae sensu lato based on rbcL sequence with incomplete taxon coverage: the limits to inference. BMC data. Annals of the Missouri Botanical Garden 80 : 631 – 660 . Evolutionary Biology 10 : 155 . Muellner , A. N. , D. D. Vassiliades , and S. S. Renner . 2007 . Placing Sauquet , H. , J. A. Doyle , T. Scharaschkin , T. Borsch , K. W. Hilu , Biebersteiniaceae, a herbaceous clade of Sapindales, in a temporal and L. W. Chatrou , and A. L. Thomas . 2003 . Phylogenetic analysis geographic context. Plant Systematics and Evolution 266 : 233 – 252 . of Magnoliales and Myristicaceae based on multiple data sets: im- Nickrent , D. L. , A. Blarer , Y.-L. Qiu , R. Vidal-Russell , and F. E. plications for character evolution. Botanical Journal of the Linnean Anderson . 2004 . Phylogenetic inference in Raffl esiales: The in- Society 142 : 125 – 186 . fl uence of rate heterogeneity and horizontal gene transfer. BMC Savolainen , V. , M. W. Chase , S. B. Hoot , C. M. Morton , D. E. Evolutionary Biology 4 : 40 . Soltis , C. Bayer , M. F. Fay , et al . 2000a . Phylogenetics of fl ow- Nickrent , D. L., V. Malé cot, R. Vidal-Russell, and J. P. Der . 2010 . ering plants based on combined analysis of plastid atpB and rbcL gene A revised classifi cation of Santalales. Taxon 59 : 538 – 558 . sequences. Systematic Biology 49 : 306 – 362 . Nixon , K. C. 1999 . The Parsimony Ratchet, a new method for rapid par- Savolainen , V. , M. F. Fay , D. C. Albach , A. Backlund , M. van der Bank, simony analysis. Cladistics 15 : 407 – 414 . K. M. Cameron, S. A. Johnson, et al. 2000b . Phylogeny of the eu- Nyffeler , R. , and U. Eggli . 2010 . Disintegrating Portulacaceae: A new dicots: A nearly complete familial analysis based on rbcL gene se- familial classifi cation of the suborder Portulacineae (Caryophyllales) quences. Kew Bulletin 55 : 257 – 309 . based on molecular and morphological data. Taxon 59 : 227 – 240 . Schö nenberger , J. , A. A. Anderberg , and K. J. Sytsma . 2005 . Olmstead , R. G. , K. J. Kim , R. K. Jansen , and S. J. Wagstaff . 2000 . The Molecular phylogenetics and patterns of fl oral evolution in the phylogeny of the Asteridae sensu lato based on chloroplast ndhF gene Ericales. International Journal of Plant Sciences 166 : 265 – 288 . sequences. Molecular Phylogenetics and Evolution 16 : 96 – 112 . Sch ö nenberger , J. , M. von Balthazar , and K. J. Sytsma . 2010 . Olmstead , R. G. , H. Michaels , K. M. Scott , and J. D. Palmer . 1992 . Diversity and evolution of fl oral structure among early diverging lin- Monophyly of the Asteridae and identifi cation of their major lin- eages in the Ericales. Philosophical Transactions of the Royal Society eages inferred from DNA sequences of rbc L. Annals of the Missouri of London, B, Biological Sciences 365 : 437 – 448 . Botanical Garden 79 : 249 – 265 . Sikes , D. S. , and P. O. Lewis. 2001. PAUPRat: A tool to implement par- Panero , J. L. , and V. A. Funk . 2008 . The value of sampling anomalous simony and likelihood ratchet searches using PAUP* [online computer taxa in phylogenetic studies: Major clades of the Asteraceae revealed. program]. Website http://www.ucalgary.ca/~dsikes/software2.htm . Molecular Phylogenetics and Evolution 47 : 757 – 782 . Simmons , M. P. , C. C. Clevinger , V. Savolainen , R. H. Archer , S. Potter , D. , T. Eriksson , R. C. Evans , S.-H. Oh , J. Smedmark , D. Mathews , and J. J. Doyle . 2001a . Phylogeny of the Celastraceae Morgan , M. Kerr , et al . 2007 . Phylogeny and classifi cation of inferred from phytochrome B gene sequence and morphology. Rosaceae. Plant Systematics and Evolution 266 : 5 – 43 . American Journal of Botany 88 : 313 – 325 . Potter , D. , F. Gao , P. E. Bortiri , S.-H. Oh , and S. Baggett . 2002 . Simmons , M. P. , V. Savolainen , C. C. Clevinger , R. H. Archer , and J. Phylogenetic relationships in Rosaceae inferred from chloroplast I. Davis . 2001b . Phylogeny of the Celastraceae inferred from 26S matK and trnL-trnF nucleotide sequence data. Plant Systematics and nuclear ribosomal DNA, phytochrome B, rbcL, atpB , and morphol- Evolution 231 : 77 – 89 . ogy. Molecular Phylogenetics and Evolution 19 : 353 – 366 . 728 American Journal of Botany [Vol. 98

Smith , S. A. , J. M. Beaulieu , and M. J. Donoghue. 2009 . Mega- historical and biogeographical analysis of Myrtaceae, Vochysiaceae, phylogeny approach for comparative biology: An alternative to and relatives in the southern hemisphere. International Journal of supertree and supermatrix approaches. BMC Evolutionary Biology Plant Sciences 165 ( 4 supplement ): S85 – S105 . 9 : 37 . Sytsma , K. J. , J. Morawetz , J. C. Pires , M. Nepokroeff , E. Conti , Smith , S. A. , and C. W. Dunn . 2008 . Phyutility: A phyloinformatics M. Zjhra , J. C. Hall , and M. W. Chase . 2002 . Urticalean rosids: tool for trees, alignments and molecular data. Bioinformatics 24 : Circumscription, rosid ancestry, and phylogenetics based on rbcL, trnL-F , 715 – 716 . and ndhF sequences. American Journal of Botany 89 : 1531 – 1546 . Soltis , D. E. , M. A. Gitzendanner , and P. S. Soltis . 2007 . A 567-taxon Talavera , G. , and J. Castresana . 2007 . Improvement of phylogenies data set for angiosperms: The challenges posed by Bayesian analy- after removing divergent and ambiguously aligned blocks from pro- ses of large data sets. International Journal of Plant Sciences 168 : tein sequence alignments. Systematic Biology 56 : 564 – 577 . 137 – 157 . Tank , D. C. , and M. J. Donoghue . 2010 . Phylogeny and phylogenetic Soltis , D. E. , and P. S. Soltis . 1997 . Phylogenetic relationships in nomenclature of the Campanulidae based on an expanded sample of Saxifragaceae sensu lato: A comparison of topologies based on genes and taxa. Systematic Botany 35 : 425 – 441 . 18S rDNA and rbcL sequences. American Journal of Botany 84 : Von Balthazar , M. , and J. Schö nenberger . 2009 . Floral struc- 504 – 522 . ture and organization in Platanaceae. International Journal of Plant Soltis , P. S. , D. E. Soltis , and M. W. Chase . 1999 . Angiosperm phy- Sciences 170 : 210 – 225 . logeny inferred from multiple genes as a tool for comparative biology. Wang , H. , M. J. Moore , P. S. Soltis , C. D. Bell , S. Brockington , R. Nature 402 : 402 – 404 . Alexandre , C. C. Davis , et al . 2009 . Rosid radiation and the rapid Soltis , D. E. , P. S. Soltis , M. W. Chase , M. E. Mort , D. C. Albach , rise of angiosperm-dominated forests. Proceedings of the National M. Zanis , V. Savolainen , et al. 2000 . Angiosperm phylogeny in- Academy of Sciences, USA 106 : 3853 – 3858 . ferred from 18S rDNA, rbcL , and atpB sequences. Botanical Journal Wang , W. , A.-M. Lu , Y. Ren , M. E. Endress , and Z.-D. Chen . 2009 . of the Linnean Society 133 : 381 – 461 . Phylogeny and classifi cation of Ranunculales: Evidence from four Soltis , D. E. , P. S. Soltis , T. G. Collier , and M. L. Edgerton . 1991 . molecular loci and morphological data. Perspectives in Plant Ecology, Chloroplast DNA variation within and among genera of the Heuchera Evolution and Systematics 11 : 81 – 110 . group: cpDNA evidence for intergeneric chloroplast transfer and the Winkworth , R. C. , J. Lundberg , and M. J. Donoghue . 2008 . Toward paraphyly of Mitella and Heuchera. American Journal of Botany 78 : a resolution of campanulid phylogeny, with a special reference to the 1091 – 1112 . placement of Dipsacales. Taxon 57 : 53 – 65 . Soltis , D. E. , P. S. Soltis , P. K. Endress , and M. W. Chase . 2005 . Wojciechowski , M. F. , M. Lavin , and M. J. Sanderson . 2004 . A phy- Phylogeny and evolution of the angiosperms. Sinauer, Sunderland, logeny of legumes (Leguminosae) based on analysis of the plastid Massachusetts, USA. matK gene resolves many well-supported subclades within the family. Soltis , D. E. , P. S. Soltis , D. L. Nickrent , L. A. Johnson , W. J. Hahn , American Journal of Botany 91 : 1846 – 1862 . S. B. Hoot , J. A. Sweere , et al . 1997 . Angiosperm phylogeny in- Wurdack , K. J. , and C. C. Davis . 2009 . Malpighiales phylogenetics: ferred from 18S ribosomal DNA sequences. Annals of the Missouri Gaining ground on one of the most recalcitrant clades in the angio- Botanical Garden 84 : 1 – 49 . sperm tree of life. American Journal of Botany 96 : 1551 – 1570 . Stamatakis , A. 2006 . RAxML-VI-HPC: Maximum likelihood-based Zanis , M. J. , D. E. Soltis , P. S. Soltis , S. Mathews , and M. J. Donoghue . phylogenetic analyses with thousands of taxa and mixed models. 2002 . The root of the angiosperms revisited. Proceedings of the Bioinformatics 22 : 2688 – 2690 . National Academy of Sciences, USA 99 : 6848 – 6853 . Stamatakis , A. , P. Hoover , and J. Rougemont . 2008 . A fast bootstrap- Zhang , L.-B. , and S. S. Renner . 2003 . The deepest splits in Chloran- ping algorithm for the RAxML web-servers. Systematic Biology 57 : thaceae as resolved by chloroplast sequences. International Journal of 758 – 771 . Plant Sciences 164 ( 5 Supplement ): S383 – S392 . Steel , M. , and M. J. Sanderson . 2010 . Characterizing phylogenetically Zhang , L.-B. , and M. P. Simmons . 2006 . Phylogeny and delimitation of decisive taxon coverage. Applied Mathematics Letters 23 : 82 – 86 . the Celastrales inferred from nuclear and plastid genes. Systematic Stevens , P. F. 2001 onward. Angiosperm Phylogeny Website, version 9, Botany 31 : 122 – 137 . June 2008 [and more or less continuously updated since]. Website Zhang , L.-B. , M. P. Simmons , A. Kocyan , and S. S. Renner . 2006 . http://www.mobot.org/MOBOT/research/APweb/ . Phylogeny of the Cucurbitales based on DNA sequences of nine Swofford , D. L. 2002 . PAUP*: Phylogenetic analysis using parsi- loci from three genomes: Implications for morphological and sex- mony (*and other methods), version 4b10. Sinauer, Sunderland, ual system evolution. Molecular Phylogenetics and Evolution 3 9 : Massachusetts, USA. 305 – 322 . Sytsma , K. J. 1994 . DNA extraction from recalcitrant plants: Long, pure, Zhang , W.-H. , Z.-D. Chen , J. H. Li , H.-B. Chen , and Y.-C. Tang . 2003 . and simple? In R. P. Adams, J. S. Miller, E. M. Golenberg, and J. E. Phylogeny of the Dipsacales s.l. based on chloroplast trnL-F and ndhF Adams [eds.], Conservation of plant genes II: Intellectual property sequences. Molecular Phylogenetics and Evolution 26 : 176 – 189 . rights and DNA utilization, 69 – 81. Missouri Botanical Garden, St. Zhu , X.-Y. , M. W. Chase , Y.-L. Qiu , H.-Z. Kong , D. L. Dilcher , J.-H. Louis, Missouri, USA. Li , and Z. D. Chen . 2007 . Mitochondrial matR sequences help to Sytsma , K. J. , A. Litt , M. L. Zjhra , J. C. Pires , M. Nepokroeff , E. Conti , resolve deep phylogenetic relationships in rosids. BMC Evolutionary J. Walker , and P. G. Wilson . 2004 . Clades, clocks, and continents: Biology 7 : 217 . April 2011] Soltis et al. — Angiosperm phylogeny 729

Appendix 1. Phylocode description of Superrosidae

Superrosidae W. S. Judd, D. E. Soltis, and P. S. Soltis clade was also well supported in both the 12-gene ML analysis of H. Wang et al. (2009 ; fi g. 1 ), a study that focused only on relationships Defi nition within the Rosidae and thus included many fewer sampled taxa, The most inclusive crown clade that contains Rosa cinnamomea L. 1753 especially among nonrosid taxa, and the 83-gene ML and MP analyses ( Rosidae ) but not Aster amellus L. 1753 (Asteridae ). This is a branch- of Moore et al. (2010; fi g. 1), which also were based on a much smaller modifi ed node-based defi nition. Abbreviated defi nition: > ∇ Rosa array of sampled taxa. The clade Superrosidae was recovered (but cinnamomea L. 1753 ~ Aster amellus L. 1753. without strong support) in all most parsimonious trees in the three- gene analyses of Soltis et al. (2000 ; see fi gs. 1(B) , 5, and 6) and in Etymology some of the most parsimonious trees resulting from the analysis of atpB From the Latin super (above, over, or on top) and Rosidae, a converted clade and rbcL sequences in Savolainen et al. (2000a) . The clade received name based on Rosa , the Latin name for rose (and probably originally weak support in the ML analyses of angiosperms based on fi ve genes from the Greek, rhodon ), in reference to the fact that the Superrosidae is (Burleigh et al., 2009, see fi g. 3 and the “ full tree” included with their intended to apply to a crown clade “ above Rosidae . ” online supplemental information). Therefore, although the existence of Reference phylogeny the Superrosidae was suspected, and this clade was informally named in Moore et al. (2010), we have not previously provided a formal name The primary reference phylogeny is this paper, see Figs. 1 and 2; see also H. for this clade. Only with the present results in hand are we suffi ciently Wang et al. (2009 ) and Moore et al. (2010) . Rosa cinnamomea is used as confi dent that Superrosidae represents a well-supported clade and thus a specifi er because it is the type species of Rosaceae (and thus Rosidae , is in need of a formal scientifi c name. under ICBN, which forms part of this clade name); its close relationship to the species of Rosaceae included in the primary reference phylogeny (i.e., There is some disagreement, however, among recent phylogenetic analyses Spiraea betulifolia , S. vanhouttei, and Prunus persica) is supported by the regarding the position of Dilleniaceae. In the 17-gene analysis (reported analyses of Evans et al. (2000) and Potter et al. (2002, 2007 ), in which here), Dilleniaceae (represented by Tetracera , Hibbertia , and Dillenia ) are Rosaceae are shown to be monophyletic. Similarly, Aster amellus is used as strongly placed as sister to a clade comprising Santalales, Caryophyllales , a specifi er because it is the type species of Asteraceae (and thus Asteridae , Berberidopsidales, and Asteridae , while in an analysis of complete plastid under ICBN); its close relationships to the species of Asteraceae included genome sequence data (Moore et al., 2010) they are placed as sister to a in the primary reference phylogeny (i.e., Barnadesia arborea , Gerbera Saxifragales + Rosidae clade. Finally, in an analysis of inverted repeat jamesonii , Echinops bannaticus , Tragopogon dubius , T. porrifolius , (IR) sequences (see Moore et al., 2010), Dilleniaceae are placed as the Cichorium intybus , Lactuca sativa , Tagetes erecta , Guizotia abyssinica , sister group to a large clade comprising Rosidae , Saxifragales, Asteridae , and Helianthus annuus ) is supported by the series of phylogenetic studies Berberidopsidales, Caryophyllales , and Santalales. The use of branch- of Asteraceae presented in Funk et al. (2009) . modifi ed, node-based defi nitions for both Superrosidae and Superasteridae accommodates placement of Dilleniaceae in either Superrosidae or Composition Superasteridae , or positioned as the sister taxon to a Superrosidae + Rosidae (incl. Vitaceae), Saxifragales, and possibly Dilleniaceae. Superasteridae clade, within the Pentapetalae (which in turn is nested within the Gunneridae ). Diagnostic apomorphies No non-DNA synapomorphies are known. The essential feature of our concept of Superrosidae is its inclusion of everything that is more closely related to Rosidae (i.e., Fabidae , Malvidae , Vitaceae) Synonyms than to Asteridae (i.e., Lamiidae , Campanulidae, Ericales, Cornales). This There are no synonymous scientifi c names but the informal name “ super- is the feature that we have tried to capture in our defi nition. Furthermore, rosids ” was used for this clade in Moore et al. (2010) , while the name when this defi nition is used in conjunction with our reciprocal defi nition “ superrosids ” is used in version 9 of Stevens (2001 , onwards). of Superasteridae (see Appendix 2), it ensures that Superrosidae and Superasteridae are always mutually exclusive, regardless of the placement Comments of Dilleniaceae. This is a recently discovered clade that is strongly supported by the extensive molecular analyses reported in this study, although we note that this

Appendix 2. PhyloCode description of Superasteridae

Superasteridae W. S. Judd, D. E. Soltis, & P. S. Soltis is used as a specifi er because it is the type species of Asteraceae (and thus Asteridae, under ICBN, which forms part of this clade name); its Defi nition close relationship to the species of Asteraceae included in the reference The most inclusive crown clade that contains Aster amellus L. 1753 ( Asteridae ) phylogeny (i.e., Barnadesia arborea , Gerbera jamesonii , Echinops but not Rosa cinnamomea L. 1753 (Rosidae ). This is a branch-modifi ed bannaticus , Tragopogon dubius , T. porrifolius , Cichorium intybus , node-based defi nition. Abbreviated defi nition: > ∇ Aster amellus L. 1753 Lactuca sativa , Tagetes erecta , Guizotia abyssinica, and Helianthus ~ Rosa cinnamomea L. 1753. annuus ) is supported by the series of phylogenetic studies of Asteraceae Rosa cinnamomea Etymology presented in Funk et al. (2009) . Similarly, is used as a specifi er because it is the type species of Rosaceae (and thus Rosidae , From the Latin super (above, over, or on top) and Asteridae, a converted clade under ICBN). The close relationship of Rosa cinnamomea to the species name based on Aster (derived from the Latin, aster , meaning star, so of Rosaceae included in the reference phylogeny (i.e., Spiraea betulifolia , called because of the form of the radiate fl oral heads of these plants), in S. vanhouttei, and Prunus persica) is supported by the analyses of Evans reference to the fact that Superasteridae is intended to refer to a crown et al. (2000) and Potter et al. (2002 , 2007 ), in which Rosaceae are shown clade “ above Asteridae . ” to be monophyletic. Reference phylogeny Composition The reference phylogeny is this paper; see Figs. 1 and 2. See also Burleigh Santalales, Berberidopsidales, Caryophyllales , Asteridae , and possibly et al. (2009, including online supplemental fi les 6 and 7). Aster amellus Dilleniaceae. 730 American Journal of Botany

Diagnostic apomorphies There is some disagreement among recent phylogenetic analyses regarding No non-DNA synapomorphies are known. the position of Dilleniaceae. In the 17-gene analysis (reported here) Dilleniaceae (represented by Tetracera , Hibbertia , and Dillenia ) are Synonyms strongly placed as sister to a clade comprising Santalales, Caryophyllales , There are no synonymous scientifi c names, but the informal names “ super- Berberidopsidales, and Asteridae , while in an analysis of complete plastid asterids ” ( Moore et al., 2010 ) and “ superasterids ” Stevens (2001 , onward, genome sequence data (Moore et al., 2010 they are placed as sister to a version 9) have been used for this clade. Saxifragales + Rosidae clade. Finally, in an analysis of IR sequences (see Moore et al., unpublished) Dilleniaceae are placed as the sister group to a Comments large clade comprising Rosidae , Saxifragales, Asteridae , Berberidopsidales, This is a recently discovered clade that is strongly supported by the extensive Caryophyllales , and Santalales. The use of branch-modifi ed node-based molecular analyses reported in this study, although we note that this defi nitions for both Superasteridae and Superrosidae accommodates clade was also strongly supported in the 83-gene ML and MP analyses placement of Dilleniaceae in either Superasteridae or Superrosidae , or of Moore et al. (2010; see fi g. 1), which were based on a much smaller positioned as the sister taxon to a Superasteridae + Superrosidae clade, array of sampled taxa. The Superasteridae received weak support in the within the Pentapetalae (which is in turn nested within the Gunneridae ). ML analyses of angiosperms based on fi ve genes (Burleigh et al., 2009, see fi g. 3 and their “ full tree” included with the online supplemental The essential feature of our concept of Superasteridae that we have tried to capture information). Therefore, although the existence of the Superasteridae in our defi nition is its inclusion of everything that is closer to Asteridae was suspected, and this clade was informally named by Moore et al. (i.e., Lamiidae , Campanulidae , Ericales, Cornales) than to Rosidae (i.e., (2010) , we have not previously provided a formal name for this clade. Fabidae , Malvidae, Vitaceae). Furthermore, when this defi nition is used in Only with the present results in hand are we suffi ciently confi dent that conjunction with our reciprocal defi nition of Superrosidae (see Appendix Superasteridae represents a well-supported clade, and thus is in need of 1), it ensures that Superasteridae and Superrosidae are always mutually a formal scientifi c name. exclusive, regardless of the placement of Dilleniaceae.